Vol. 13/ Núm. 3 2026 pág. 1990
https://doi.org/10.69639/arandu.v13i3.2518
ChatGPT Voice and Role-Play Activities on English Oral
Communicative Competence: A Pilot Study with Secondary
Learners
ChatGPT Voice y actividades de juego de roles para fortalecer la competencia
comunicativa oral en inglés: Un estudio piloto con estudiantes de educación secundaria
Gabriela Jacqueline Galeas Arboleda
ggaleas@uteq.edu.ec
https://orcid.org/0000-0001-9759-5474
Universidad Técnica Estatal de Quevedo
Quevedo – Ecuador
David Enrique Montenegro Montenegro
dmontenegro@yachaytech.edu.ec
https://orcid.org/0009-0009-2549-7256
Yachay Tech University
Urcuquí – Ecuador
Joel Sebastián Chevez Mendaño
https://orcid.org/0009-0001-2907-1190
jchevezm@uteq.edu.ec
Universidad Técnica Estatal de Quevedo
Quevedo – Ecuador
Artículo recibido: 10 julio 2026- Aceptado para publicación:16 agosto 2026
Conflictos de intereses: Ninguno que declarar.
ABSTRACT
The rapid implementation of AI (Artificial Intelligence) technology has gained the attention of
many educators to enhance complex EFL skills, such as speaking, especially in contexts with
limited opportunities for language practice. This pilot study inspects the effects of the use of
ChatGPT's voice feature to enhance oral presentations (role-plays). A group of secondary EFL
learners in a public school participated in a four-week intervention, engaging in daily practice
with ChatGPT voice. In each English session, participants engaged in a role-play practice with
the AI feature. A speaking rubric with five sub-skills was designed and used to assess students'
performance each week, and a semi-structured interview was conducted to collect specific
participants’ perceptions. ChatGPT was employed to evaluate student performance, with the
exception of interaction/expressiveness, to get more objective results. To examine students'
performance across the four weeks and to address the first research question regarding their
progression in oral communicative competence, a repeated-measures analysis of variance
(ANOVA) was conducted to observe statistically significant changes over time. Qualitative data
were analyzed using thematic analysis. Quantitative data indicated that ChatGPT Voice
effectively enhanced EFL students´ oral skills in combination with role-plays; however, no
Vol. 13/ Núm. 3 2026 pág. 1991
significant differences in gains were observed among subskills. On the other hand, qualitative
analysis portrayed positive perceptions of the AI tool for speaking practice in alignment with
previous research and major considerations of its usage. The study offers important insights for
language teachers and researchers into the use of AI technologies and how they complement
traditional instruction and assessment.
Keywords: AI, EFL oral communication, role-play, secondary education
RESUMEN
La rápida incorporación de tecnologías de inteligencia artificial (IA) ha despertado el interés de
docentes por fortalecer habilidades complejas del inglés como lengua extranjera (EFL),
especialmente la expresión oral, en contextos con oportunidades limitadas para practicar el
idioma. Este estudio piloto examinó los efectos del uso de la función de voz de ChatGPT
(ChatGPT Voice) para mejorar las presentaciones orales mediante actividades de juego de roles.
Participó un grupo de estudiantes de educación secundaria de una institución pública en una
intervención de cuatro semanas. Durante este periodo, los estudiantes practicaron diariamente con
ChatGPT Voice y realizaron actividades de juego de roles en las sesiones de inglés. El desempeño
oral se evaluó semanalmente mediante una rúbrica diseñada para el estudio, que incluyó cinco
subhabilidades. Además, se realizaron entrevistas semiestructuradas para conocer las
percepciones de algunos participantes. ChatGPT se empleó como herramienta de evaluación,
excepto para la dimensión de interacción/expresividad, con el propósito de favorecer una
valoración más objetiva. Para analizar los cambios en el desempeño oral a lo largo de las cuatro
semanas, se aplicó un análisis de varianza de medidas repetidas (ANOVA), mientras que los datos
cualitativos se examinaron mediante análisis temático. Los resultados cuantitativos mostraron que
ChatGPT Voice, combinado con actividades de juego de roles, contribuyó significativamente a
mejorar las habilidades orales de los estudiantes, aunque no se encontraron diferencias
significativas entre las ganancias de las distintas subhabilidades. El análisis cualitativo reveló
percepciones positivas y consideraciones relevantes sobre el uso de esta herramienta. El estudio
aporta evidencia sobre el potencial de la IA para complementar la enseñanza y evaluación
tradicionales de lenguas.
Palabras claves: inteligencia artificial, comunicación oral de Inglés como lengua
extranjera, juego de roles, educación secundaria
Todo el contenido de la Revista Científica Internacional Arandu UTIC publicado en este sitio está disponible bajo
licencia Creative Commons Atribution 4.0 International.
significant differences in gains were observed among subskills. On the other hand, qualitative
analysis portrayed positive perceptions of the AI tool for speaking practice in alignment with
previous research and major considerations of its usage. The study offers important insights for
language teachers and researchers into the use of AI technologies and how they complement
traditional instruction and assessment.
Keywords: AI, EFL oral communication, role-play, secondary education
RESUMEN
La rápida incorporación de tecnologías de inteligencia artificial (IA) ha despertado el interés de
docentes por fortalecer habilidades complejas del inglés como lengua extranjera (EFL),
especialmente la expresión oral, en contextos con oportunidades limitadas para practicar el
idioma. Este estudio piloto examinó los efectos del uso de la función de voz de ChatGPT
(ChatGPT Voice) para mejorar las presentaciones orales mediante actividades de juego de roles.
Participó un grupo de estudiantes de educación secundaria de una institución pública en una
intervención de cuatro semanas. Durante este periodo, los estudiantes practicaron diariamente con
ChatGPT Voice y realizaron actividades de juego de roles en las sesiones de inglés. El desempeño
oral se evaluó semanalmente mediante una rúbrica diseñada para el estudio, que incluyó cinco
subhabilidades. Además, se realizaron entrevistas semiestructuradas para conocer las
percepciones de algunos participantes. ChatGPT se empleó como herramienta de evaluación,
excepto para la dimensión de interacción/expresividad, con el propósito de favorecer una
valoración más objetiva. Para analizar los cambios en el desempeño oral a lo largo de las cuatro
semanas, se aplicó un análisis de varianza de medidas repetidas (ANOVA), mientras que los datos
cualitativos se examinaron mediante análisis temático. Los resultados cuantitativos mostraron que
ChatGPT Voice, combinado con actividades de juego de roles, contribuyó significativamente a
mejorar las habilidades orales de los estudiantes, aunque no se encontraron diferencias
significativas entre las ganancias de las distintas subhabilidades. El análisis cualitativo reveló
percepciones positivas y consideraciones relevantes sobre el uso de esta herramienta. El estudio
aporta evidencia sobre el potencial de la IA para complementar la enseñanza y evaluación
tradicionales de lenguas.
Palabras claves: inteligencia artificial, comunicación oral de Inglés como lengua
extranjera, juego de roles, educación secundaria
Todo el contenido de la Revista Científica Internacional Arandu UTIC publicado en este sitio está disponible bajo
licencia Creative Commons Atribution 4.0 International.
Vol. 13/ Núm. 3 2026 pág. 1992
INTRODUCTION
The acquisition of English as a foreign language remains a central focus in secondary
education, where oral fluency is often considered one of the most demanding skills to develop.
Many learners struggle to express themselves fluently and confidently in English due to limited
opportunities for authentic communicative practice, which frequently leads to anxiety, low
motivation, and reduced oral performance (Fernández-García & Fonseca-Mora, 2019). As
highlighted by Sosa-López and Mora (2022), speaking anxiety has a direct impact on learners’
fluency, accuracy, and overall communicative complexity, making it one of the most critical
challenges in English language learning.
In response to speaking difficulties, artificial intelligence (AI) has emerged as a promising
support for foreign language education. Recent advances in AI-powered tools, such as
conversational chatbots, provide learners with interactive and adaptive environments that
simulate real-life communication. According to Koç and Savaş (2024), the integration of AI
chatbots in English language learning enhances motivation, increases learner engagement, and
offers opportunities for individualized practice, thereby reinforcing speaking skills in more
natural and dynamic ways.
Empirical studies indicate that AI-mediated interactions not only improve fluency but
also contribute to a broader range of speaking competencies, including pronunciation, lexical
development, grammatical accuracy, and willingness to communicate (Fathi et al., 2024; Nuñez
et al., 2025). These findings suggest that AI tools, when integrated with pedagogical strategies
such as role-playing, can significantly enhance oral language learning outcomes, provided
appropriate pedagogical considerations are in place (Kim & Park, 2023).
There is a considerable body of studies on AI technologies used in EFL/ESL speaking
contexts, predominantly in Asian countries such as China or Indonesia (Xing, 2025; Fauzi et al.,
2025). For instance, a similar study was carried out by Huang (2024), who investigated how, by
using voice prompts, students can leverage the potential of ChatGPT to gain meaningful feedback
on their oral performance and strengthen their pronunciation skills. However, this research was
developed in a culturally different background. Furthermore, little is known about AI technologies
to enhance speaking skills in secondary learners for oral presentation, specifically role-plays.
Most studies have been carried out in higher education (Deng & Jamaludin, 2026).
Building on this perspective, the present study examines the effect of ChatGPT Voice
Chat on the oral competence of secondary school students in role-play activities. The research is
conducted in Quevedo, a coastal city characterized by public educational contexts in which
opportunities for oral English practice are often limited and largely confined to the classroom.
Within the broader context of Ecuador, English as a Foreign Language (EFL) instruction has
gained increasing importance in recent years; however, persistent challenges remain, particularly
INTRODUCTION
The acquisition of English as a foreign language remains a central focus in secondary
education, where oral fluency is often considered one of the most demanding skills to develop.
Many learners struggle to express themselves fluently and confidently in English due to limited
opportunities for authentic communicative practice, which frequently leads to anxiety, low
motivation, and reduced oral performance (Fernández-García & Fonseca-Mora, 2019). As
highlighted by Sosa-López and Mora (2022), speaking anxiety has a direct impact on learners’
fluency, accuracy, and overall communicative complexity, making it one of the most critical
challenges in English language learning.
In response to speaking difficulties, artificial intelligence (AI) has emerged as a promising
support for foreign language education. Recent advances in AI-powered tools, such as
conversational chatbots, provide learners with interactive and adaptive environments that
simulate real-life communication. According to Koç and Savaş (2024), the integration of AI
chatbots in English language learning enhances motivation, increases learner engagement, and
offers opportunities for individualized practice, thereby reinforcing speaking skills in more
natural and dynamic ways.
Empirical studies indicate that AI-mediated interactions not only improve fluency but
also contribute to a broader range of speaking competencies, including pronunciation, lexical
development, grammatical accuracy, and willingness to communicate (Fathi et al., 2024; Nuñez
et al., 2025). These findings suggest that AI tools, when integrated with pedagogical strategies
such as role-playing, can significantly enhance oral language learning outcomes, provided
appropriate pedagogical considerations are in place (Kim & Park, 2023).
There is a considerable body of studies on AI technologies used in EFL/ESL speaking
contexts, predominantly in Asian countries such as China or Indonesia (Xing, 2025; Fauzi et al.,
2025). For instance, a similar study was carried out by Huang (2024), who investigated how, by
using voice prompts, students can leverage the potential of ChatGPT to gain meaningful feedback
on their oral performance and strengthen their pronunciation skills. However, this research was
developed in a culturally different background. Furthermore, little is known about AI technologies
to enhance speaking skills in secondary learners for oral presentation, specifically role-plays.
Most studies have been carried out in higher education (Deng & Jamaludin, 2026).
Building on this perspective, the present study examines the effect of ChatGPT Voice
Chat on the oral competence of secondary school students in role-play activities. The research is
conducted in Quevedo, a coastal city characterized by public educational contexts in which
opportunities for oral English practice are often limited and largely confined to the classroom.
Within the broader context of Ecuador, English as a Foreign Language (EFL) instruction has
gained increasing importance in recent years; however, persistent challenges remain, particularly
Vol. 13/ Núm. 3 2026 pág. 1993
the limited exposure to authentic communicative situations. This underscores the need to explore
innovative tools that can effectively support the development of linguistic competence.
Over a four-week intervention, students engaged in role-play activities guided by teacher-
designed prompts that simulated real-life communicative scenarios (e.g., everyday interactions,
academic situations, and social contexts). These activities were mediated through ChatGPT Voice
Chat, enabling continuous interaction with an artificial intelligence system capable of providing
immediate and adaptive feedback. This approach aimed not only to increase learners’ exposure
to the target language but also to create a more dynamic, accessible, and low-anxiety learning
environment.
The primary purpose of this study is to evaluate the extent to which the use of ChatGPT
Voice Chat + role-play contributes to improvements in key components of oral competence,
including fluency, pronunciation, vocabulary use, grammatical accuracy, and speaking
confidence. Additionally, the study explores students’ perceptions of artificial intelligence as a
supportive tool in their English learning process, considering dimensions such as motivation,
learner autonomy, and the reduction of language anxiety.
It is important to note that, although there has been growing international interest in the
use of artificial intelligence in language education, there is still limited empirical evidence in the
Latin American context—and particularly in Ecuador—regarding the implementation of tools
such as ChatGPT Voice for the development of oral skills in secondary education. Furthermore,
Chat GPT Voice and role-plays contribute to an under-researched area of speaking with
voice-enabled Gen-AI in secondary contexts (Lo et al, 2024).
Therefore, this study seeks to address this gap in the literature by providing context-
specific evidence that contributes to a better understanding of the potential of these technologies
in similar educational settings.
The purpose of this pilot study is to explore the potential of ChatGPT Voice as an
innovative digital tool to enhance English oral proficiency among secondary school students
within a short intervention period. To achieve this aim, the following question was set:
• To what extent does the use of ChatGPT Voice Chat improve the speaking skills of
secondary students during a four-week intervention?
Secondary question:
• What measurable changes are observed in students’ pronunciation, fluency, vocabulary,
and grammar after practicing with ChatGPT Voice Chat?
• How do secondary students perceive the use of ChatGPT Voice for practicing role-plays?
LITERATURE REVIEW
Speaking Skills in EFL
Among the four core language skills—reading, writing, listening, and speaking—
speaking is widely recognized as the most difficult to develop, despite being essential for effective
the limited exposure to authentic communicative situations. This underscores the need to explore
innovative tools that can effectively support the development of linguistic competence.
Over a four-week intervention, students engaged in role-play activities guided by teacher-
designed prompts that simulated real-life communicative scenarios (e.g., everyday interactions,
academic situations, and social contexts). These activities were mediated through ChatGPT Voice
Chat, enabling continuous interaction with an artificial intelligence system capable of providing
immediate and adaptive feedback. This approach aimed not only to increase learners’ exposure
to the target language but also to create a more dynamic, accessible, and low-anxiety learning
environment.
The primary purpose of this study is to evaluate the extent to which the use of ChatGPT
Voice Chat + role-play contributes to improvements in key components of oral competence,
including fluency, pronunciation, vocabulary use, grammatical accuracy, and speaking
confidence. Additionally, the study explores students’ perceptions of artificial intelligence as a
supportive tool in their English learning process, considering dimensions such as motivation,
learner autonomy, and the reduction of language anxiety.
It is important to note that, although there has been growing international interest in the
use of artificial intelligence in language education, there is still limited empirical evidence in the
Latin American context—and particularly in Ecuador—regarding the implementation of tools
such as ChatGPT Voice for the development of oral skills in secondary education. Furthermore,
Chat GPT Voice and role-plays contribute to an under-researched area of speaking with
voice-enabled Gen-AI in secondary contexts (Lo et al, 2024).
Therefore, this study seeks to address this gap in the literature by providing context-
specific evidence that contributes to a better understanding of the potential of these technologies
in similar educational settings.
The purpose of this pilot study is to explore the potential of ChatGPT Voice as an
innovative digital tool to enhance English oral proficiency among secondary school students
within a short intervention period. To achieve this aim, the following question was set:
• To what extent does the use of ChatGPT Voice Chat improve the speaking skills of
secondary students during a four-week intervention?
Secondary question:
• What measurable changes are observed in students’ pronunciation, fluency, vocabulary,
and grammar after practicing with ChatGPT Voice Chat?
• How do secondary students perceive the use of ChatGPT Voice for practicing role-plays?
LITERATURE REVIEW
Speaking Skills in EFL
Among the four core language skills—reading, writing, listening, and speaking—
speaking is widely recognized as the most difficult to develop, despite being essential for effective
Vol. 13/ Núm. 3 2026 pág. 1994
communication and real-life interaction (Rao, 2019). Learners often struggle to demonstrate their
speaking abilities due to multiple challenges, including speaking anxiety, limited vocabulary and
grammatical knowledge, pronunciation difficulties, and insufficient opportunities for oral practice
(Du et al., 2025; Aziz & Kashinathan, 2021). These factors frequently result in low participation
and hinder the development of communicative competence in EFL contexts.
A primary goal of learning a foreign language is the ability to communicate; however,
many students lack confidence when attempting to improve their speaking skills (Fernández-
García & Fonseca-Mora, 2019). In response to these challenges, recent research highlights the
growing role of artificial intelligence in enhancing oral proficiency. AI-based tools help reduce
foreign language anxiety (FLA), promote self-regulated learning, and provide adaptive
instruction. By creating low-pressure conversational environments, these tools enable learners to
practice speaking without fear of judgment, thereby fostering both confidence and fluency.
Moreover, their flexible and explicit design allows students to monitor their progress through
continuous oral feedback, which further supports the development of learner autonomy (Nhan et
al., 2025). In this sense, the integration of technology into the development of oral skills facilitates
the design of student-centered activities that actively promote autonomous learning (Cuevas-
Montero, 2021).
According to the Common European Framework of Reference for Languages, speaking
competence involves the ability to interact orally, express ideas, and engage effectively in
conversations across proficiency levels (Council of Europe, 2020). Nevertheless, the development
of speaking skills in EFL classrooms remains constrained by various factors, such as students’
anxiety during interaction and difficulties in applying grammatical knowledge (Tapia & Vega,
2024). In the Ecuadorian context, these challenges are further compounded by systemic issues,
including limited instructional time, an emphasis on grammar and writing over oral
communication, a shortage of qualified English teachers, and low student motivation. Together,
these barriers continue to limit opportunities for meaningful speaking practice in the classroom
(Rodriguez & Baquerizo, 2025).
GenAI and EFL Speaking
Generative Artificial Intelligence (GenAI) refers to artificial intelligence systems capable
of generating new content, such as text, images, or audio, by learning patterns from large datasets
through advanced machine learning models (Lim, 2023). Furthermore, large language models
such as ChatGPT represent a significant advancement in artificial intelligence because they can
generate human-like responses and support interactive learning processes in educational contexts
(Kasneci et al., 2023). Despite the benefits of generative AI in education, several challenges have
been identified. For example, large language models may generate inaccurate information, raise
ethical concerns, and encourage excessive reliance on technology if not used critically in
educational contexts (Kasneci et al., 2023).
communication and real-life interaction (Rao, 2019). Learners often struggle to demonstrate their
speaking abilities due to multiple challenges, including speaking anxiety, limited vocabulary and
grammatical knowledge, pronunciation difficulties, and insufficient opportunities for oral practice
(Du et al., 2025; Aziz & Kashinathan, 2021). These factors frequently result in low participation
and hinder the development of communicative competence in EFL contexts.
A primary goal of learning a foreign language is the ability to communicate; however,
many students lack confidence when attempting to improve their speaking skills (Fernández-
García & Fonseca-Mora, 2019). In response to these challenges, recent research highlights the
growing role of artificial intelligence in enhancing oral proficiency. AI-based tools help reduce
foreign language anxiety (FLA), promote self-regulated learning, and provide adaptive
instruction. By creating low-pressure conversational environments, these tools enable learners to
practice speaking without fear of judgment, thereby fostering both confidence and fluency.
Moreover, their flexible and explicit design allows students to monitor their progress through
continuous oral feedback, which further supports the development of learner autonomy (Nhan et
al., 2025). In this sense, the integration of technology into the development of oral skills facilitates
the design of student-centered activities that actively promote autonomous learning (Cuevas-
Montero, 2021).
According to the Common European Framework of Reference for Languages, speaking
competence involves the ability to interact orally, express ideas, and engage effectively in
conversations across proficiency levels (Council of Europe, 2020). Nevertheless, the development
of speaking skills in EFL classrooms remains constrained by various factors, such as students’
anxiety during interaction and difficulties in applying grammatical knowledge (Tapia & Vega,
2024). In the Ecuadorian context, these challenges are further compounded by systemic issues,
including limited instructional time, an emphasis on grammar and writing over oral
communication, a shortage of qualified English teachers, and low student motivation. Together,
these barriers continue to limit opportunities for meaningful speaking practice in the classroom
(Rodriguez & Baquerizo, 2025).
GenAI and EFL Speaking
Generative Artificial Intelligence (GenAI) refers to artificial intelligence systems capable
of generating new content, such as text, images, or audio, by learning patterns from large datasets
through advanced machine learning models (Lim, 2023). Furthermore, large language models
such as ChatGPT represent a significant advancement in artificial intelligence because they can
generate human-like responses and support interactive learning processes in educational contexts
(Kasneci et al., 2023). Despite the benefits of generative AI in education, several challenges have
been identified. For example, large language models may generate inaccurate information, raise
ethical concerns, and encourage excessive reliance on technology if not used critically in
educational contexts (Kasneci et al., 2023).
Vol. 13/ Núm. 3 2026 pág. 1995
On the other hand, according to Deng and Jamaludin (2026), just a few studies focus on
GenAI in speaking skills, even though AI systems offer students meaningful opportunities to
engage in real-life language use while receiving prompt and constructive feedback (Mageira et
al., 2022; Park, 2023). ChatGPT Voice proved highly effective in improving learners’ oral
proficiency, showing observable progress in pronunciation accuracy, fluency, vocabulary range,
and grammatical precision. These improvements were largely associated with sustained AI-
supported practice and the provision of immediate corrective feedback (Nuñez et al., 2025).
Recently, a study found that AI-based dialogue systems fulfill an important communicative
function in EFL learning at the university level, as students report that these systems reduce
speaking anxiety, improve vocabulary retention, and provide instant feedback (Zhai & Wibowo,
2023).
The integration of AI-assisted tools has shown considerable potential for enhancing
English language learning, particularly in the development of speaking skills. Evidence suggests
that regular and sustained use—rather than simple access—plays a crucial role in achieving
positive learning outcomes, especially in underserved or rural settings. In these contexts, AI
technologies offer a scalable, adaptive, and learner-centered solution to support the development
of oral fluency. Furthermore, such tools are both pedagogically sound and feasible for
implementation in real classroom environments (Alenezi & Alenezi, 2025). Studies also indicate
that learners make noticeable progress in key aspects of speaking, including intonation, stress,
and fluency, while benefiting from increased opportunities for practice both within and beyond
the classroom. In addition, AI chatbots promote self-regulated learning by providing adaptive
support and personalized feedback, thereby enriching the overall learning experience.
Moreover, this technology helps reduce language anxiety and speaking-related
apprehension, creating supportive environments that encourage active participation and
confidence building (Klímová & Seraj, 2023). Their ease of use and accessibility make them
suitable for informal learning contexts, enabling students to engage in meaningful communication
at any time without requiring advanced technical skills. Short, guided interactions are particularly
effective for automated assessment, while strategies such as Socratic questioning contribute to
the development of critical thinking skills in EFL learners. Importantly, students with lower levels
of proficiency tend to benefit the most, demonstrating gains in confidence, vocabulary
development, self-assessment, and oral presentation abilities. Additionally, research on tools such
as ChatGPT emphasizes the relevance of personalized feedback and careful instructional design,
offering valuable insights for optimizing oral communication practice in EFL contexts (Cha et
al., 2024).
ChatGPT Voice Chat
The adoption and integration of AI tools across various sectors have rapidly increased
since the launch of ChatGPT in 2022. This trend has also reached the educational field, where
On the other hand, according to Deng and Jamaludin (2026), just a few studies focus on
GenAI in speaking skills, even though AI systems offer students meaningful opportunities to
engage in real-life language use while receiving prompt and constructive feedback (Mageira et
al., 2022; Park, 2023). ChatGPT Voice proved highly effective in improving learners’ oral
proficiency, showing observable progress in pronunciation accuracy, fluency, vocabulary range,
and grammatical precision. These improvements were largely associated with sustained AI-
supported practice and the provision of immediate corrective feedback (Nuñez et al., 2025).
Recently, a study found that AI-based dialogue systems fulfill an important communicative
function in EFL learning at the university level, as students report that these systems reduce
speaking anxiety, improve vocabulary retention, and provide instant feedback (Zhai & Wibowo,
2023).
The integration of AI-assisted tools has shown considerable potential for enhancing
English language learning, particularly in the development of speaking skills. Evidence suggests
that regular and sustained use—rather than simple access—plays a crucial role in achieving
positive learning outcomes, especially in underserved or rural settings. In these contexts, AI
technologies offer a scalable, adaptive, and learner-centered solution to support the development
of oral fluency. Furthermore, such tools are both pedagogically sound and feasible for
implementation in real classroom environments (Alenezi & Alenezi, 2025). Studies also indicate
that learners make noticeable progress in key aspects of speaking, including intonation, stress,
and fluency, while benefiting from increased opportunities for practice both within and beyond
the classroom. In addition, AI chatbots promote self-regulated learning by providing adaptive
support and personalized feedback, thereby enriching the overall learning experience.
Moreover, this technology helps reduce language anxiety and speaking-related
apprehension, creating supportive environments that encourage active participation and
confidence building (Klímová & Seraj, 2023). Their ease of use and accessibility make them
suitable for informal learning contexts, enabling students to engage in meaningful communication
at any time without requiring advanced technical skills. Short, guided interactions are particularly
effective for automated assessment, while strategies such as Socratic questioning contribute to
the development of critical thinking skills in EFL learners. Importantly, students with lower levels
of proficiency tend to benefit the most, demonstrating gains in confidence, vocabulary
development, self-assessment, and oral presentation abilities. Additionally, research on tools such
as ChatGPT emphasizes the relevance of personalized feedback and careful instructional design,
offering valuable insights for optimizing oral communication practice in EFL contexts (Cha et
al., 2024).
ChatGPT Voice Chat
The adoption and integration of AI tools across various sectors have rapidly increased
since the launch of ChatGPT in 2022. This trend has also reached the educational field, where
Vol. 13/ Núm. 3 2026 pág. 1996
these technologies are becoming increasingly present in everyday teaching and learning practices
(Bustamante Salinas et al., 2025). Recently, more and more instructors are incorporating Chatbots
in speaking instructions (Klímová & Seraj, 2023; Tai & Chen, 2024) as they offer opportunities
for learners to engage in meaningful language input and productive output within a low-stress
environment (Jeon et al., 2023). Pau et al. (2026) emphasize that, despite technical challenges,
ChatGPT stands out significantly, excelling in the ability to adapt and be evaluated in real-time,
as well as in integration with virtual reality (VR) technologies. Its use has increased significantly
since 2023 (Xing, 2025).
Conversational AI tools such as ChatGPT enable learners to interact with spoken dialogue
systems, allowing them to practice the target language through simulated conversations. These
tools provide opportunities for students to receive individualized feedback while communicating
in a low-anxiety environment, which can support the development of their speaking skills
(Ericsson & Johansson, 2023). AI-mediated speaking activities allow learners to practice oral
communication skills in interactive learning environments. Through these technologies, students
can engage in repeated speaking practice, which may contribute to improvements in fluency,
pronunciation, and grammatical accuracy in second-language learning contexts (Fathi et al.,
2024). Studies also show that AI chatbots can enhance speaking outcomes by increasing learners’
confidence, engagement, and motivation during oral language practice. By providing
opportunities for constant interaction, these tools may encourage students to participate more
actively in speaking activities (Du & Daniel, 2024).
METHODOLOGY
This pilot study adopted a mixed-method exploratory research design and aims to provide
preliminary evidence of ChatGPT Voice for oral English development among secondary school
learners. A mixed- method approach is particularly well-suited for language learning studies, as
it enables a comprehensive understanding of both outcomes and learner experiences (Wei, 2022;
Niglas et al., 2021).
Participants
Given the exploratory nature of the intervention, fifteen high school students, aged 12–
15, were initially nominated to participate in the study. The group included students of both
genders, allowing for a more diverse range of experiences and perceptions within the learning
process. Subsequently, an oral diagnostic interview was administered to the entire group to
determine the final participants. This evaluation assessed key components of oral competence,
including pronunciation, fluency, vocabulary use, and grammatical accuracy.
The total scores were converted into percentages, and seven students who achieved 70%
or higher were selected to form the final sample. This selection was based on intentional sampling,
considering their academic performance in English and their relevance to the analysis of oral
these technologies are becoming increasingly present in everyday teaching and learning practices
(Bustamante Salinas et al., 2025). Recently, more and more instructors are incorporating Chatbots
in speaking instructions (Klímová & Seraj, 2023; Tai & Chen, 2024) as they offer opportunities
for learners to engage in meaningful language input and productive output within a low-stress
environment (Jeon et al., 2023). Pau et al. (2026) emphasize that, despite technical challenges,
ChatGPT stands out significantly, excelling in the ability to adapt and be evaluated in real-time,
as well as in integration with virtual reality (VR) technologies. Its use has increased significantly
since 2023 (Xing, 2025).
Conversational AI tools such as ChatGPT enable learners to interact with spoken dialogue
systems, allowing them to practice the target language through simulated conversations. These
tools provide opportunities for students to receive individualized feedback while communicating
in a low-anxiety environment, which can support the development of their speaking skills
(Ericsson & Johansson, 2023). AI-mediated speaking activities allow learners to practice oral
communication skills in interactive learning environments. Through these technologies, students
can engage in repeated speaking practice, which may contribute to improvements in fluency,
pronunciation, and grammatical accuracy in second-language learning contexts (Fathi et al.,
2024). Studies also show that AI chatbots can enhance speaking outcomes by increasing learners’
confidence, engagement, and motivation during oral language practice. By providing
opportunities for constant interaction, these tools may encourage students to participate more
actively in speaking activities (Du & Daniel, 2024).
METHODOLOGY
This pilot study adopted a mixed-method exploratory research design and aims to provide
preliminary evidence of ChatGPT Voice for oral English development among secondary school
learners. A mixed- method approach is particularly well-suited for language learning studies, as
it enables a comprehensive understanding of both outcomes and learner experiences (Wei, 2022;
Niglas et al., 2021).
Participants
Given the exploratory nature of the intervention, fifteen high school students, aged 12–
15, were initially nominated to participate in the study. The group included students of both
genders, allowing for a more diverse range of experiences and perceptions within the learning
process. Subsequently, an oral diagnostic interview was administered to the entire group to
determine the final participants. This evaluation assessed key components of oral competence,
including pronunciation, fluency, vocabulary use, and grammatical accuracy.
The total scores were converted into percentages, and seven students who achieved 70%
or higher were selected to form the final sample. This selection was based on intentional sampling,
considering their academic performance in English and their relevance to the analysis of oral
Vol. 13/ Núm. 3 2026 pág. 1997
competence development. In this way, the process ensured alignment between the study’s
objectives and the characteristics of the participants. The informed consents were approved by
the legal guardians of participants who must be students under 18 years old. Pseudonyms were
used to protect the identities of participants.
The study was conducted in a public educational setting characterized by limited
resources and few opportunities for authentic English interaction outside of the classroom. These
conditions make the development of oral skills a significant challenge for students, reinforcing
the relevance of implementing technology-mediated pedagogical interventions. It should be
mentioned that the students had previously experienced using ChatGPT, but not for oral English
practice.
Instruments
The techniques used for data collection in this research are as follows:
• A speaking rubric. The speaking rubric was developed by the researchers and informed
by established frameworks for assessing oral proficiency (Brown & Abeywickrama, 2019;
Luoma, 2004). The instrument assessed five key dimensions of speaking performance:
pronunciation, fluency, vocabulary, grammar, and interaction. To ensure content validity, the
rubric was reviewed by an expert in English language teaching and assessment, who examined
the relevance, clarity, and appropriateness of each criterion. The expert’s recommendations were
incorporated into the final version of the instrument before its implementation. Additionally, a
second evaluator independently assessed a subset of the speaking performances to enhance
scoring reliability and reduce potential evaluator bias (McNamara, 1996). The rubric was
employed to evaluate students’ performance in a role-play task based on a fashion show scenario,
which served as the final project of the intervention. The primary purpose was to assess their
speaking skills, considering all aspects involved in oral communication, and to monitor their
progress over the weeks.
Tabla 1
Rubric for Evaluating Speaking in a Role Play
Criteria 2 pts (Excellent) 1 pt (Acceptable) 0 pts (Deficient)
Pronunciation
Clear and
understandable
pronunciation; few
difficulties in
comprehension.
Mostly
understandable,
though with some
mistakes.
Very difficult to
understand; frequent
errors.
Fluency
Speaks naturally,
continuously, with no
Some pauses or
fillers, but meaning is
maintained.
Many pauses, rigid
reading, or lack of
fluency.
competence development. In this way, the process ensured alignment between the study’s
objectives and the characteristics of the participants. The informed consents were approved by
the legal guardians of participants who must be students under 18 years old. Pseudonyms were
used to protect the identities of participants.
The study was conducted in a public educational setting characterized by limited
resources and few opportunities for authentic English interaction outside of the classroom. These
conditions make the development of oral skills a significant challenge for students, reinforcing
the relevance of implementing technology-mediated pedagogical interventions. It should be
mentioned that the students had previously experienced using ChatGPT, but not for oral English
practice.
Instruments
The techniques used for data collection in this research are as follows:
• A speaking rubric. The speaking rubric was developed by the researchers and informed
by established frameworks for assessing oral proficiency (Brown & Abeywickrama, 2019;
Luoma, 2004). The instrument assessed five key dimensions of speaking performance:
pronunciation, fluency, vocabulary, grammar, and interaction. To ensure content validity, the
rubric was reviewed by an expert in English language teaching and assessment, who examined
the relevance, clarity, and appropriateness of each criterion. The expert’s recommendations were
incorporated into the final version of the instrument before its implementation. Additionally, a
second evaluator independently assessed a subset of the speaking performances to enhance
scoring reliability and reduce potential evaluator bias (McNamara, 1996). The rubric was
employed to evaluate students’ performance in a role-play task based on a fashion show scenario,
which served as the final project of the intervention. The primary purpose was to assess their
speaking skills, considering all aspects involved in oral communication, and to monitor their
progress over the weeks.
Tabla 1
Rubric for Evaluating Speaking in a Role Play
Criteria 2 pts (Excellent) 1 pt (Acceptable) 0 pts (Deficient)
Pronunciation
Clear and
understandable
pronunciation; few
difficulties in
comprehension.
Mostly
understandable,
though with some
mistakes.
Very difficult to
understand; frequent
errors.
Fluency
Speaks naturally,
continuously, with no
Some pauses or
fillers, but meaning is
maintained.
Many pauses, rigid
reading, or lack of
fluency.
Vol. 13/ Núm. 3 2026 pág. 1998
long pauses or overuse
of notes.
Vocabulary
Varied and appropriate
vocabulary for the role
play context.
Limited or repetitive
vocabulary, but
understandable.
Very basic
vocabulary, frequent
incorrect usage.
Grammar
Correct sentences, few
errors that do not affect
meaning.
Some grammatical
errors, but overall
understandable.
Many errors that
make comprehension
difficult.
Interaction /
Expressiveness
Actively participates,
responds naturally, uses
gestures, and
appropriate intonation.
Participates but with
little expressiveness
or minimal
interaction.
Does not interact or
depend fully on notes,
without
expressiveness.
Source: Own production
Note: This table shows the criteria for evaluating a speaking role play. Scores range from 0 to 2
points: 2 = Excellent, 1 = Acceptable, 0 = Deficient.
• Semi-structured interview. A semi-structured interview was employed to explore
students’ perceptions, experiences, and evaluations of the learning process while allowing
flexibility in the interaction and the opportunity to probe emerging ideas (Kallio et al., 2016). This
type of interview is particularly suitable for educational research because it enables the collection
of rich and detailed data while maintaining a clear focus on the research objectives (Creswell &
Creswell, 2018; Cohen et al., 2018). Following the four-week intervention, each participant
completed a short semi-structured interview consisting of two open-ended questions. Although
limited in number, the questions were carefully designed to elicit reflections on the use of
ChatGPT Voice Chat as a learning support tool and its perceived influence on confidence and
speaking development. The use of open-ended questions encouraged participants to provide
spontaneous and detailed responses, generating insights into affective and cognitive aspects of
language learning, such as motivation, anxiety, and self-perceived progress. The interview data
complemented the quantitative findings obtained through the speaking rubric, contributing to a
more comprehensive understanding of the intervention’s impact. To facilitate participants’
freedom of expression and ensure the accuracy of their responses, the interviews were conducted
in Spanish, their native language.
Intervention
During the initial phase of the intervention, a mechanical practice based on pre-written
dialogues was implemented to familiarize students with the dynamics of the activity. In this sense,
during the first week, participants worked with dialogues previously designed and provided by
the researchers, corresponding to a role-playing game that included two presenters (Angela and
long pauses or overuse
of notes.
Vocabulary
Varied and appropriate
vocabulary for the role
play context.
Limited or repetitive
vocabulary, but
understandable.
Very basic
vocabulary, frequent
incorrect usage.
Grammar
Correct sentences, few
errors that do not affect
meaning.
Some grammatical
errors, but overall
understandable.
Many errors that
make comprehension
difficult.
Interaction /
Expressiveness
Actively participates,
responds naturally, uses
gestures, and
appropriate intonation.
Participates but with
little expressiveness
or minimal
interaction.
Does not interact or
depend fully on notes,
without
expressiveness.
Source: Own production
Note: This table shows the criteria for evaluating a speaking role play. Scores range from 0 to 2
points: 2 = Excellent, 1 = Acceptable, 0 = Deficient.
• Semi-structured interview. A semi-structured interview was employed to explore
students’ perceptions, experiences, and evaluations of the learning process while allowing
flexibility in the interaction and the opportunity to probe emerging ideas (Kallio et al., 2016). This
type of interview is particularly suitable for educational research because it enables the collection
of rich and detailed data while maintaining a clear focus on the research objectives (Creswell &
Creswell, 2018; Cohen et al., 2018). Following the four-week intervention, each participant
completed a short semi-structured interview consisting of two open-ended questions. Although
limited in number, the questions were carefully designed to elicit reflections on the use of
ChatGPT Voice Chat as a learning support tool and its perceived influence on confidence and
speaking development. The use of open-ended questions encouraged participants to provide
spontaneous and detailed responses, generating insights into affective and cognitive aspects of
language learning, such as motivation, anxiety, and self-perceived progress. The interview data
complemented the quantitative findings obtained through the speaking rubric, contributing to a
more comprehensive understanding of the intervention’s impact. To facilitate participants’
freedom of expression and ensure the accuracy of their responses, the interviews were conducted
in Spanish, their native language.
Intervention
During the initial phase of the intervention, a mechanical practice based on pre-written
dialogues was implemented to familiarize students with the dynamics of the activity. In this sense,
during the first week, participants worked with dialogues previously designed and provided by
the researchers, corresponding to a role-playing game that included two presenters (Angela and
Vol. 13/ Núm. 3 2026 pág. 1999
Deymar). Students were required to memorize their interventions and practice with ChatGPT,
focused solely on checking whether they could reproduce the dialogue correctly. At this stage,
the chatbot played a fundamentally mechanical role: listening, comparing, and validating the
accuracy of the memorized lines.
At a later stage of the intervention, a progressive practice based on oral interaction was
promoted through the voice chat function. From the second week to the fourth, students began
interacting with ChatGPT, producing increasingly complex dialogues. The purpose of this stage
was to observe gradual development in key aspects such as grammar, vocabulary, fluency, and
pronunciation. In this context, ChatGPT took on the role of an active interlocutor, allowing for a
more authentic practice and providing feedback that facilitated the identification of progress in
participants' oral competence. Figure 1 shows the two main stages and the virtual assistance role.
In order to structure the interaction during practice activities clearly and consistently, the
researchers designed a specific prompt that guided the use of the tool. The statement provided
was as follows: "First of all, let’s have a conversation using your voice chat feature with this RPG.
You will say all the sentences of the presenter, both Angela and Deymar. Let’s do the whole
conversation at the same time. We started with voice chat, and after that, I’ll ask you for ratings
and feedback." This prompt made it possible to establish an organized dynamic, ensuring that the
interaction developed in a coherent way and aligned with the objectives of the activity.
This instruction also made it easier for ChatGPT to take an active role in the process,
simultaneously representing both dialogue presenters. In this way, the system not only functioned
as an interlocutor but also as a linguistic model that guided the structure of the conversation,
maintaining continuity of exchange and offering students a clear reference for language use. This
mediation contributed to a more controlled and structured practice environment, in which
participants could focus on oral production. The design of the prompt also incorporated a later
evaluation phase, in which the tool was required to provide ratings and feedback based on student
performance. This component made it possible to immediately integrate the formative evaluation
within the same interaction, favoring processes of self-adjustment and reflection on one’s own
linguistic performance.
Deymar). Students were required to memorize their interventions and practice with ChatGPT,
focused solely on checking whether they could reproduce the dialogue correctly. At this stage,
the chatbot played a fundamentally mechanical role: listening, comparing, and validating the
accuracy of the memorized lines.
At a later stage of the intervention, a progressive practice based on oral interaction was
promoted through the voice chat function. From the second week to the fourth, students began
interacting with ChatGPT, producing increasingly complex dialogues. The purpose of this stage
was to observe gradual development in key aspects such as grammar, vocabulary, fluency, and
pronunciation. In this context, ChatGPT took on the role of an active interlocutor, allowing for a
more authentic practice and providing feedback that facilitated the identification of progress in
participants' oral competence. Figure 1 shows the two main stages and the virtual assistance role.
In order to structure the interaction during practice activities clearly and consistently, the
researchers designed a specific prompt that guided the use of the tool. The statement provided
was as follows: "First of all, let’s have a conversation using your voice chat feature with this RPG.
You will say all the sentences of the presenter, both Angela and Deymar. Let’s do the whole
conversation at the same time. We started with voice chat, and after that, I’ll ask you for ratings
and feedback." This prompt made it possible to establish an organized dynamic, ensuring that the
interaction developed in a coherent way and aligned with the objectives of the activity.
This instruction also made it easier for ChatGPT to take an active role in the process,
simultaneously representing both dialogue presenters. In this way, the system not only functioned
as an interlocutor but also as a linguistic model that guided the structure of the conversation,
maintaining continuity of exchange and offering students a clear reference for language use. This
mediation contributed to a more controlled and structured practice environment, in which
participants could focus on oral production. The design of the prompt also incorporated a later
evaluation phase, in which the tool was required to provide ratings and feedback based on student
performance. This component made it possible to immediately integrate the formative evaluation
within the same interaction, favoring processes of self-adjustment and reflection on one’s own
linguistic performance.
Vol. 13/ Núm. 3 2026 pág. 2000
Figura 1
Intervention Process
Source: Own production
It is important to note that the original presenters (the characters of the role-playing game)
were not included as participants in the study because they only read the scripted lines. They were
not involved in the process of learning or improvement. Their role was purely performative.
During the actual practice sessions, ChatGPT acted as a presenter, involving each student.
Performance evaluation and data analysis
Performance was assessed at two levels. The first level was assessed by Automatic
Evaluation by ChatGPT. The virtual assistant provided feedback on grammar, vocabulary,
fluency, and pronunciation. However, the researchers did not use the chatbot’s automatic scoring
for interaction or expressiveness, because the system produced random values and could not
accurately detect emotions, acting skills, or expressiveness. The second level was human
evaluation (done by researchers). The researchers evaluated the criteria related to interaction and
expressiveness.
Data analysis retrieved from the criteria was carried out using the SPSS Statistics
program. Descriptive statistics (means and standard deviations) were calculated to examine
students' performance across the four weeks and to address the first research question regarding
their progression in oral communicative competence. To determine whether the observed changes
over time were statistically significant, a repeated-measures analysis of variance (ANOVA) was
Figura 1
Intervention Process
Source: Own production
It is important to note that the original presenters (the characters of the role-playing game)
were not included as participants in the study because they only read the scripted lines. They were
not involved in the process of learning or improvement. Their role was purely performative.
During the actual practice sessions, ChatGPT acted as a presenter, involving each student.
Performance evaluation and data analysis
Performance was assessed at two levels. The first level was assessed by Automatic
Evaluation by ChatGPT. The virtual assistant provided feedback on grammar, vocabulary,
fluency, and pronunciation. However, the researchers did not use the chatbot’s automatic scoring
for interaction or expressiveness, because the system produced random values and could not
accurately detect emotions, acting skills, or expressiveness. The second level was human
evaluation (done by researchers). The researchers evaluated the criteria related to interaction and
expressiveness.
Data analysis retrieved from the criteria was carried out using the SPSS Statistics
program. Descriptive statistics (means and standard deviations) were calculated to examine
students' performance across the four weeks and to address the first research question regarding
their progression in oral communicative competence. To determine whether the observed changes
over time were statistically significant, a repeated-measures analysis of variance (ANOVA) was
Vol. 13/ Núm. 3 2026 pág. 2001
conducted. Qualitative data obtained from the open-ended interview questions were analyzed
using thematic analysis. Responses were coded, categorized into themes, and supported with
representative excerpts from participants' comments. Pseudonyms were employed to protect
participants' identities.
RESULTS
Quantitative Results
Over the four weeks of implementation, students demonstrated progressive improvement
in their oral English performance. Table 1 shows that in Week 1, during the initial diagnosis and
first practices, most participants showed willingness to participate and attempted to answer in
English; however, their performance was weak, with very short answers, frequent pauses, and
limited vocabulary, resulting in an average score of (M=4,59). By Week 2, the first improvements
became evident as students began producing longer answers and using simple connectors such as
“because “and “and”. Despite this progress, many pauses, grammar errors, and limited
expressiveness persisted, with an average score of (M=6,5). During Week 3, students showed
steady progress, delivering more natural speech, organizing ideas more clearly, and incorporating
more connectors, which allowed them to give more detailed answers. Although some grammar
and pronunciation mistakes remained, comprehension was much clearer, and scores improved to
(M=7,44). Finally, in Week 4, during the last practices and the role-play presentation, students
reached their best performance, providing fluent, detailed, and confident responses with clear and
coherent communication. Only minor errors were observed, and their average scores increased to
(M=8,67), marking a significant improvement compared to the first week. Students’ weekly
scores over four weeks and their total progress. Total progress is calculated as the difference
between the Week 1 and Week 4 scores (see Table 1).
Tabla 2
Students’ Progress Over Four Weeks
N Mean Std. Deviation Minimum Maximum
Week 1 7 4.59 .20 4.30 4.80
Week 2 7 6.50 .22 6.20 6.80
Week 3 7 7.44 .17 7.20 7.70
Week 4 7 8.67 .11 8.50 8.80
To determine whether students' speaking performance improved over time, a Friedman
test was performed on the weekly rubric scores. The analysis revealed a statistically significant
difference among the four measurement points (p < .001). This finding demonstrated that students'
speaking performance improved significantly throughout the intervention. The progressive
conducted. Qualitative data obtained from the open-ended interview questions were analyzed
using thematic analysis. Responses were coded, categorized into themes, and supported with
representative excerpts from participants' comments. Pseudonyms were employed to protect
participants' identities.
RESULTS
Quantitative Results
Over the four weeks of implementation, students demonstrated progressive improvement
in their oral English performance. Table 1 shows that in Week 1, during the initial diagnosis and
first practices, most participants showed willingness to participate and attempted to answer in
English; however, their performance was weak, with very short answers, frequent pauses, and
limited vocabulary, resulting in an average score of (M=4,59). By Week 2, the first improvements
became evident as students began producing longer answers and using simple connectors such as
“because “and “and”. Despite this progress, many pauses, grammar errors, and limited
expressiveness persisted, with an average score of (M=6,5). During Week 3, students showed
steady progress, delivering more natural speech, organizing ideas more clearly, and incorporating
more connectors, which allowed them to give more detailed answers. Although some grammar
and pronunciation mistakes remained, comprehension was much clearer, and scores improved to
(M=7,44). Finally, in Week 4, during the last practices and the role-play presentation, students
reached their best performance, providing fluent, detailed, and confident responses with clear and
coherent communication. Only minor errors were observed, and their average scores increased to
(M=8,67), marking a significant improvement compared to the first week. Students’ weekly
scores over four weeks and their total progress. Total progress is calculated as the difference
between the Week 1 and Week 4 scores (see Table 1).
Tabla 2
Students’ Progress Over Four Weeks
N Mean Std. Deviation Minimum Maximum
Week 1 7 4.59 .20 4.30 4.80
Week 2 7 6.50 .22 6.20 6.80
Week 3 7 7.44 .17 7.20 7.70
Week 4 7 8.67 .11 8.50 8.80
To determine whether students' speaking performance improved over time, a Friedman
test was performed on the weekly rubric scores. The analysis revealed a statistically significant
difference among the four measurement points (p < .001). This finding demonstrated that students'
speaking performance improved significantly throughout the intervention. The progressive
Vol. 13/ Núm. 3 2026 pág. 2002
increase in scores across the four weeks suggests that regular practice with ChatGPT Voice may
have contributed positively to the development of learners' oral skills.
Table 2 illustrates the weekly progress of the entire group in relation to each assessment
criterion. The table shows that participants predominantly outperformed in vocabulary from
(M=0,83) in week 1 to (M=1,86). Followed by fluency with (M=0,67) in the initial week to
(M=1,67) in the final week of practice, and pronunciation with (M=0,77) in the first week and
(M=1,71). Then the grammar criterion went from (M=0,97) to (M=1.64). Finally,
interaction/expressiveness started at (M=1,30) and reached (M=1,79) at the end of the
intervention.
Tabla 3
Progression over the weeks by each criterion
Week 1 Week 2 Week 3 Week 4
Criteria Mean Mean Mean Mean
Pronunciation 0.77 1.29 1.86 1.71
Fluency 0.67 1.14 1.43 1.67
Vocabulary 0.83 1.46 1.43 1.86
Grammar 0.97 1.14 1.49 1.64
Interaction/ Expressiveness 1.30 1.47 1.24 1.79
Total 4.8 6.7 7.6 8.8
Note: The data presents the Median scores of each criterion.
A repeated-measures ANOVA in Table 3 revealed a statistically significant effect of
Week on students’ speaking performance, F(3. 18) = 4157.33, p < .001, η² = .999, indicating a
substantial and consistent improvement across the four-week intervention. In contrast, no
significant main effect of Subskill was found, F(4, 24) = 0.62, p = .656, η² = .09, suggesting that
students performed similarly across pronunciation, fluency, vocabulary, grammar, and
interaction. The Week × Subskill interaction was also not significant, F(12, 72) = 1.00, p = .457,
η² = .14, indicating that all speaking subskills improved at a comparable rate over time.
Tabla 3
Repeated-measures ANOVA for speaking performance
Effect F df p Partial η²
Week 4157.33 3. 18 < .001 .999
Subskill 0.62 4. 24 .656 .093
Week × Subskill 1.00 12. 72 .457 .143
Figure 2 illustrates the progression of students’ speaking subskills across the four-week
intervention. Overall, all subskills exhibited an upward trend, indicating consistent improvement
increase in scores across the four weeks suggests that regular practice with ChatGPT Voice may
have contributed positively to the development of learners' oral skills.
Table 2 illustrates the weekly progress of the entire group in relation to each assessment
criterion. The table shows that participants predominantly outperformed in vocabulary from
(M=0,83) in week 1 to (M=1,86). Followed by fluency with (M=0,67) in the initial week to
(M=1,67) in the final week of practice, and pronunciation with (M=0,77) in the first week and
(M=1,71). Then the grammar criterion went from (M=0,97) to (M=1.64). Finally,
interaction/expressiveness started at (M=1,30) and reached (M=1,79) at the end of the
intervention.
Tabla 3
Progression over the weeks by each criterion
Week 1 Week 2 Week 3 Week 4
Criteria Mean Mean Mean Mean
Pronunciation 0.77 1.29 1.86 1.71
Fluency 0.67 1.14 1.43 1.67
Vocabulary 0.83 1.46 1.43 1.86
Grammar 0.97 1.14 1.49 1.64
Interaction/ Expressiveness 1.30 1.47 1.24 1.79
Total 4.8 6.7 7.6 8.8
Note: The data presents the Median scores of each criterion.
A repeated-measures ANOVA in Table 3 revealed a statistically significant effect of
Week on students’ speaking performance, F(3. 18) = 4157.33, p < .001, η² = .999, indicating a
substantial and consistent improvement across the four-week intervention. In contrast, no
significant main effect of Subskill was found, F(4, 24) = 0.62, p = .656, η² = .09, suggesting that
students performed similarly across pronunciation, fluency, vocabulary, grammar, and
interaction. The Week × Subskill interaction was also not significant, F(12, 72) = 1.00, p = .457,
η² = .14, indicating that all speaking subskills improved at a comparable rate over time.
Tabla 3
Repeated-measures ANOVA for speaking performance
Effect F df p Partial η²
Week 4157.33 3. 18 < .001 .999
Subskill 0.62 4. 24 .656 .093
Week × Subskill 1.00 12. 72 .457 .143
Figure 2 illustrates the progression of students’ speaking subskills across the four-week
intervention. Overall, all subskills exhibited an upward trend, indicating consistent improvement
Vol. 13/ Núm. 3 2026 pág. 2003
in oral performance. Fluency showed the most stable linear development, while pronunciation
demonstrated rapid initial gains followed by slight stabilization in Week 4. Vocabulary and
grammar showed steady improvement with minor fluctuations, whereas
interaction/expressiveness displayed a more variable pattern, with noticeable improvement
emerging in the final week. These findings suggest that while all speaking components benefited
from the intervention, each subskill developed at a different rate and with distinct trajectories.
Figure 2
Changes in Learners’ Speaking Performance by Subskill
Qualitative Results
After completing four weeks of practice, two open-ended questions were asked of each
student to understand their perceptions and feelings about the process and the use of ChatGPT as
a support tool. The qualitative findings reveal overwhelmingly positive perceptions of ChatGPT
Voice Chat as a supportive tool for enhancing English-speaking skills and confidence.
Participants consistently described emotional shifts from anxiety and nervousness to confidence
and comfort throughout the intervention.
Several students emphasized the reduction of anxiety in speaking, which coincides with
the affective filter hypothesis in the acquisition of second languages (Krashen, 1982). For
example, Ana reported feeling "more confident practicing without pressure", highlighting
improvements in both pronunciation and fluency, which suggests that an environment without
immediate assessment favors oral production. In contrast, Sofia mentioned that although she felt
insecure at first, the constant repetition of interactions allowed her to gain confidence
progressively, evidencing a process of adaptation. Similarly, Elena noted that the tool made her
feel supported and calm, making it easier to organize her ideas before responding. However, other
in oral performance. Fluency showed the most stable linear development, while pronunciation
demonstrated rapid initial gains followed by slight stabilization in Week 4. Vocabulary and
grammar showed steady improvement with minor fluctuations, whereas
interaction/expressiveness displayed a more variable pattern, with noticeable improvement
emerging in the final week. These findings suggest that while all speaking components benefited
from the intervention, each subskill developed at a different rate and with distinct trajectories.
Figure 2
Changes in Learners’ Speaking Performance by Subskill
Qualitative Results
After completing four weeks of practice, two open-ended questions were asked of each
student to understand their perceptions and feelings about the process and the use of ChatGPT as
a support tool. The qualitative findings reveal overwhelmingly positive perceptions of ChatGPT
Voice Chat as a supportive tool for enhancing English-speaking skills and confidence.
Participants consistently described emotional shifts from anxiety and nervousness to confidence
and comfort throughout the intervention.
Several students emphasized the reduction of anxiety in speaking, which coincides with
the affective filter hypothesis in the acquisition of second languages (Krashen, 1982). For
example, Ana reported feeling "more confident practicing without pressure", highlighting
improvements in both pronunciation and fluency, which suggests that an environment without
immediate assessment favors oral production. In contrast, Sofia mentioned that although she felt
insecure at first, the constant repetition of interactions allowed her to gain confidence
progressively, evidencing a process of adaptation. Similarly, Elena noted that the tool made her
feel supported and calm, making it easier to organize her ideas before responding. However, other
Vol. 13/ Núm. 3 2026 pág. 2004
students indicated that sometimes the lack of explicit corrective feedback limited the
identification of specific errors. Taken together, these reflections show that the flexible and non-
critical nature of vocal interaction contributes significantly to reducing anxiety, although it also
raises the need to balance this environment with more targeted feedback instances to enhance
learning.
A recurring theme was the gradual transformation of self-perception. Sofia described
moving from being shy to feeling confident and excited about speaking in English. Julia and
Camila also reported transitioning from nervousness and anxiety to feeling prepared, comfortable,
and fluent. These emotional changes indicate that repeated exposure to voice practice fostered
both linguistic development and self-assurance.
Additionally, students highlighted gains in expressiveness and vocabulary. Isabela
mentioned feeling more relaxed and natural when speaking, while Valeria expressed pride in her
improvement, particularly in vocabulary and pronunciation. The sense of achievement and
enjoyment further reinforced their motivation to continue practicing. The findings suggest that
ChatGPT Voice Chat is perceived not only as a language practice tool but also as an emotionally
supportive environment that promotes fluency, pronunciation development, and increased
speaking confidence.
DISCUSSION
Speaking gains (Overall and subskills)
The noticeable progress in students’ ability to express themselves confidently in English
represents one of the most significant findings of this study. Participants demonstrated
improvements in pronunciation, vocabulary use, and overall communicative fluency throughout
the intervention. These findings are consistent with previous research on AI-mediated language
learning, which has reported significant gains in speaking performance through the use of
ChatGPT Voice, AI conversational agents, and chatbots, often yielding medium to large effect
sizes (Nuñez et al., 2025; Guerrero, 2026). In particular, the results of the present study closely
resemble those reported by Nuñez et al. (2025), especially regarding vocabulary development.
Notably, comparable improvements were observed despite the relatively short duration of the
intervention, which lasted only four weeks. Progress was evident in the final role-playing
practices, where students showed longer, more coherent, and natural responses compared to their
initial performances. These results are consistent with recent research on the use of artificial
intelligence tools in oral skills development, which shows significant improvements in fluency,
discursive coherence, and confidence in communication (Yan & Singh, 2026).
Speaking anxiety, engagement, and autonomy
It has also been noted that AI-mediated practice environments favor greater language
production and reduce anxiety by offering free-judgment interactions, facilitating more frequent
speaking, which resonates with Tan and Ismail's (2025) and Nhan et al. (2025) results,
students indicated that sometimes the lack of explicit corrective feedback limited the
identification of specific errors. Taken together, these reflections show that the flexible and non-
critical nature of vocal interaction contributes significantly to reducing anxiety, although it also
raises the need to balance this environment with more targeted feedback instances to enhance
learning.
A recurring theme was the gradual transformation of self-perception. Sofia described
moving from being shy to feeling confident and excited about speaking in English. Julia and
Camila also reported transitioning from nervousness and anxiety to feeling prepared, comfortable,
and fluent. These emotional changes indicate that repeated exposure to voice practice fostered
both linguistic development and self-assurance.
Additionally, students highlighted gains in expressiveness and vocabulary. Isabela
mentioned feeling more relaxed and natural when speaking, while Valeria expressed pride in her
improvement, particularly in vocabulary and pronunciation. The sense of achievement and
enjoyment further reinforced their motivation to continue practicing. The findings suggest that
ChatGPT Voice Chat is perceived not only as a language practice tool but also as an emotionally
supportive environment that promotes fluency, pronunciation development, and increased
speaking confidence.
DISCUSSION
Speaking gains (Overall and subskills)
The noticeable progress in students’ ability to express themselves confidently in English
represents one of the most significant findings of this study. Participants demonstrated
improvements in pronunciation, vocabulary use, and overall communicative fluency throughout
the intervention. These findings are consistent with previous research on AI-mediated language
learning, which has reported significant gains in speaking performance through the use of
ChatGPT Voice, AI conversational agents, and chatbots, often yielding medium to large effect
sizes (Nuñez et al., 2025; Guerrero, 2026). In particular, the results of the present study closely
resemble those reported by Nuñez et al. (2025), especially regarding vocabulary development.
Notably, comparable improvements were observed despite the relatively short duration of the
intervention, which lasted only four weeks. Progress was evident in the final role-playing
practices, where students showed longer, more coherent, and natural responses compared to their
initial performances. These results are consistent with recent research on the use of artificial
intelligence tools in oral skills development, which shows significant improvements in fluency,
discursive coherence, and confidence in communication (Yan & Singh, 2026).
Speaking anxiety, engagement, and autonomy
It has also been noted that AI-mediated practice environments favor greater language
production and reduce anxiety by offering free-judgment interactions, facilitating more frequent
speaking, which resonates with Tan and Ismail's (2025) and Nhan et al. (2025) results,
Vol. 13/ Núm. 3 2026 pág. 2005
respectively. These results are also consistent with broader evidence suggesting that AI-driven
conversational practice can strengthen fluency, lexical development, and learners’ willingness to
communicate (Fathi et al., 2024), highlighting the importance of interaction in technology-
supported oral language development. However, other studies found that anxiety was not reduced
or even slightly intensified, despite performance gains or task-specific reductions only (Huang,
2026). For instance, Shazly (2021) found that the use of conversationally enhanced AI chatbots
slightly intensified learners' FLA speech-related anxieties, which were not reduced. The reduction
of unpleasant emotions, such as anxiety, when interacting with AI technologies makes a straight
connection with a previous study done by Susoy (2026), who found a significant reduction in
speaking anxiety before AI-facilitated examinations compared to human-facilitated or neutralized
the relationship between anxiety and performance.
Student feedback further supports quantitative results, as learners reported that interacting
with ChatGPT Voice Chat provided them with consistent opportunities to practice without the
pressure commonly experienced in traditional classroom settings. This aligns with recent studies
emphasizing that repeated practice and immediate feedback are key to enhancing communicative
competence in second language learning (Sosa-López & Mora, 2022). The use of AI allowed
students to engage in meaningful and personalized tasks, enabling them to progress at their own
pace while focusing on individual weaknesses (Koç & Savaş, 2024).
Students highlighted that this approach simulated real-life communicative scenarios, which
not only reinforced their linguistic skills but also boosted their self-confidence when speaking in
public. However, while AI provides a low-pressure environment, it may not eliminate
performance-related apprehension for all learners (Huu et al., 2026). Similarly, researchers like
Wei (2023) found that AI-mediated instruction positively impacts learners’ motivation, self-
regulation, and learning outcomes, while Zhang et al. (2024) provided empirical validation for
AI-driven technology to increase students’ willingness to communicate and reduce foreign
language anxiety, especially if working in pairs on work and presentation tasks (Huang &
Kakham, 2026). Alenezi and Alenezi (2025) also reported that students expressed high levels of
satisfaction, motivation, and confidence when using the chatbot, with the overall attitude score
averaging 4.35 out of 5. Students frequently report positive attitudes, higher confidence, and a
“safe space” for experimentation with positively matches Fathi et al. (2024) and Tan and Ismail
(2025).
A central procedure in this study involved integrating an AI-based automated evaluation
system alongside human assessment to ensure fairness and balance in measuring students’
performance. This approach aligns with the findings of Susoy (2026), whose participants showed
a strong preference for human facilitation, attributing this preference to the empathy and
interpersonal connection provided by human evaluators. Although this may seem contradictory,
the evidence suggests that the most effective assessment models may integrate both AI-driven
respectively. These results are also consistent with broader evidence suggesting that AI-driven
conversational practice can strengthen fluency, lexical development, and learners’ willingness to
communicate (Fathi et al., 2024), highlighting the importance of interaction in technology-
supported oral language development. However, other studies found that anxiety was not reduced
or even slightly intensified, despite performance gains or task-specific reductions only (Huang,
2026). For instance, Shazly (2021) found that the use of conversationally enhanced AI chatbots
slightly intensified learners' FLA speech-related anxieties, which were not reduced. The reduction
of unpleasant emotions, such as anxiety, when interacting with AI technologies makes a straight
connection with a previous study done by Susoy (2026), who found a significant reduction in
speaking anxiety before AI-facilitated examinations compared to human-facilitated or neutralized
the relationship between anxiety and performance.
Student feedback further supports quantitative results, as learners reported that interacting
with ChatGPT Voice Chat provided them with consistent opportunities to practice without the
pressure commonly experienced in traditional classroom settings. This aligns with recent studies
emphasizing that repeated practice and immediate feedback are key to enhancing communicative
competence in second language learning (Sosa-López & Mora, 2022). The use of AI allowed
students to engage in meaningful and personalized tasks, enabling them to progress at their own
pace while focusing on individual weaknesses (Koç & Savaş, 2024).
Students highlighted that this approach simulated real-life communicative scenarios, which
not only reinforced their linguistic skills but also boosted their self-confidence when speaking in
public. However, while AI provides a low-pressure environment, it may not eliminate
performance-related apprehension for all learners (Huu et al., 2026). Similarly, researchers like
Wei (2023) found that AI-mediated instruction positively impacts learners’ motivation, self-
regulation, and learning outcomes, while Zhang et al. (2024) provided empirical validation for
AI-driven technology to increase students’ willingness to communicate and reduce foreign
language anxiety, especially if working in pairs on work and presentation tasks (Huang &
Kakham, 2026). Alenezi and Alenezi (2025) also reported that students expressed high levels of
satisfaction, motivation, and confidence when using the chatbot, with the overall attitude score
averaging 4.35 out of 5. Students frequently report positive attitudes, higher confidence, and a
“safe space” for experimentation with positively matches Fathi et al. (2024) and Tan and Ismail
(2025).
A central procedure in this study involved integrating an AI-based automated evaluation
system alongside human assessment to ensure fairness and balance in measuring students’
performance. This approach aligns with the findings of Susoy (2026), whose participants showed
a strong preference for human facilitation, attributing this preference to the empathy and
interpersonal connection provided by human evaluators. Although this may seem contradictory,
the evidence suggests that the most effective assessment models may integrate both AI-driven
Vol. 13/ Núm. 3 2026 pág. 2006
and human-mediated processes. Thereby capitalizing on the complementary advantages of each
modality, rather than over-relying on AI, which may undermine autonomy and critical evaluation
of alternative options due to automation (Ebadi et al., 2025).
CONCLUSION
This study showed that the systematic use of ChatGPT Voice Chat had significant positive
effects on students' oral English proficiency in high school. Through a phased implementation
over several weeks, the comparison between initial results and final evaluation based on the role
play showed clear and measurable improvements in key components of communication
competence. These developments suggest that consistent practice in interactive and simulated
environments contributes effectively to strengthening oral skills in an English as a foreign
language learning context. In addition, qualitative analysis provided a broader understanding of
the students' experience, who perceived ChatGPT Voice Chat as an accessible, innovative, and
motivating tool. In particular, they highlighted its usefulness in reducing the nervousness
associated with oral production (or level of anxiety). This environment allowed students to
experience the language more spontaneously and naturally, promoting a greater willingness to
participate and take linguistic risks. This study positions ChatGPT voice + role-play design as a
valuable contribution to enhance oral skills among secondary learners in a context where rich and
authentic exposure is limited. However, teachers may use the tool to provide safe, non-judgmental
rehearsal spaces, rather than over-reliance, a trade-off that can lead to a reduction of learner
autonomy, lower ability to communicate without AI help, and decreased motivation to generate
ideas.
Taken together, the study’s findings not only support the pedagogical value of integrating
artificial-intelligence technologies into medium-system language teaching, but they also
emphasize the importance of designing learning experiences that combine structured practice with
opportunities for meaningful interaction (role-play). Therefore, it is recommended to consider the
use of tools such as ChatGPT Voice Chat as a strategic complement to traditional English
language teaching practices (Koç & Savaş, 2024), both in the classroom and outside it, capable
of enriching teaching processes, learning, and practice while responding to the communicative
needs of students in educational contexts where English language learning is not immersive.
While the present study offers positive and relevant results, it is important to recognize
some methodological limitations that do not invalidate the findings but guide their interpretation.
First, the intervention time was four weeks, which is a relatively short period. However, this lapse
was sufficient to show significant improvements in oral production, highlighting the potential of
the tool even in short-term interventions. Secondly, the sample consisted of a small number of
students, all women. Although these characteristics limit the generalization of results to larger
and more diverse populations, it does not affect the validity of findings within the study group.
and human-mediated processes. Thereby capitalizing on the complementary advantages of each
modality, rather than over-relying on AI, which may undermine autonomy and critical evaluation
of alternative options due to automation (Ebadi et al., 2025).
CONCLUSION
This study showed that the systematic use of ChatGPT Voice Chat had significant positive
effects on students' oral English proficiency in high school. Through a phased implementation
over several weeks, the comparison between initial results and final evaluation based on the role
play showed clear and measurable improvements in key components of communication
competence. These developments suggest that consistent practice in interactive and simulated
environments contributes effectively to strengthening oral skills in an English as a foreign
language learning context. In addition, qualitative analysis provided a broader understanding of
the students' experience, who perceived ChatGPT Voice Chat as an accessible, innovative, and
motivating tool. In particular, they highlighted its usefulness in reducing the nervousness
associated with oral production (or level of anxiety). This environment allowed students to
experience the language more spontaneously and naturally, promoting a greater willingness to
participate and take linguistic risks. This study positions ChatGPT voice + role-play design as a
valuable contribution to enhance oral skills among secondary learners in a context where rich and
authentic exposure is limited. However, teachers may use the tool to provide safe, non-judgmental
rehearsal spaces, rather than over-reliance, a trade-off that can lead to a reduction of learner
autonomy, lower ability to communicate without AI help, and decreased motivation to generate
ideas.
Taken together, the study’s findings not only support the pedagogical value of integrating
artificial-intelligence technologies into medium-system language teaching, but they also
emphasize the importance of designing learning experiences that combine structured practice with
opportunities for meaningful interaction (role-play). Therefore, it is recommended to consider the
use of tools such as ChatGPT Voice Chat as a strategic complement to traditional English
language teaching practices (Koç & Savaş, 2024), both in the classroom and outside it, capable
of enriching teaching processes, learning, and practice while responding to the communicative
needs of students in educational contexts where English language learning is not immersive.
While the present study offers positive and relevant results, it is important to recognize
some methodological limitations that do not invalidate the findings but guide their interpretation.
First, the intervention time was four weeks, which is a relatively short period. However, this lapse
was sufficient to show significant improvements in oral production, highlighting the potential of
the tool even in short-term interventions. Secondly, the sample consisted of a small number of
students, all women. Although these characteristics limit the generalization of results to larger
and more diverse populations, it does not affect the validity of findings within the study group.
Vol. 13/ Núm. 3 2026 pág. 2007
Finally, the participants belonged to the same educational context, specifically a public institution.
While this homogeneity may restrict extrapolation to other environments, it also allowed for
greater control of contextual variables. In this sense, these limitations do not diminish the
relevance of the results obtained, but open opportunities for future research.
Finally, the participants belonged to the same educational context, specifically a public institution.
While this homogeneity may restrict extrapolation to other environments, it also allowed for
greater control of contextual variables. In this sense, these limitations do not diminish the
relevance of the results obtained, but open opportunities for future research.
Vol. 13/ Núm. 3 2026 pág. 2008
REFERENCES
Alenezi, A., & Alenezi, A. (2025). Evaluating the Effectiveness of Chatbot-Assisted Learning in
Enhancing English Conversational Skills Among Secondary School Students. Education
Sciences, 15(9), 1136. https://doi.org/10.3390/educsci15091136
Aziz, A. A., & Kashinathan, S. (2021). ESL learners’ challenges in speaking English in Malaysian
classroom. Development, 10(2), 983-991. https://doi.org/10.6007/ijarped/v10-i2/10355
Brown, H. D., & Abeywickrama, P. (2019). Language assessment: Principles and classroom
practices (3rd ed.). Pearson.
Bustamante Salinas, P. M., Rodríguez Maida, A. A., & Quisbert Pinedo, L. (2026). Integración
de recursos de inteligencia artificial en los procesos educativos. Revista Tecnología,
Ciencia Y Educación, (33), 139–167. https://doi.org/10.51302/tce.2026.24177
Cha, J., Han, J., Yoo, H., & Oh, A. (2024). CHOP: Integrating ChatGPT into EFL Oral
Presentation Practice. arXiv preprint arXiv:2407.07393.
https://doi.org/10.48550/arXiv.2407.07393
Cohen, L., Manion, L., & Morrison, K. (2018). Research methods in education (8th ed.).
Routledge.
Council of Europe. (2020). Common European Framework of Reference for Languages:
Learning, teaching, assessment—Companion volume. Council of Europe Publishing.
https://www.coe.int/en/web/common-european-framework-reference-languages
Creswell, J. W., & Creswell, J. D. (2018). Research design: Qualitative, quantitative, and mixed
methods approaches (5th ed.). SAGE Publications.
Cuevas-Montero, R. (2021). Desarrollo e integración de las destrezas orales en la educación a
distancia a través de podcasts colaborativos. En J. Martín Párraga & J. M. Garrido Anguita
(Eds.), De Platón al “Homo Tecnologicus”: Las humanidades en el siglo XXI (pp. 75–82).
Tirant Humanidades.
Deng, L., & Jamaludin, K. A. (2026). Roles of Generative Artificial Intelligence (GENAI) in
English as a Foreign Language (EFL) instruction: A systematic literature review. SAGE
Open, 16(1). https://doi.org/10.1177/21582440261418315
Du, T. T., Duc, L. V., The, N. H., & Dung, D. T. B. (2025). Challenges faced by EFL students at
a university in Vietnam in developing English speaking skills and pedagogical strategies
used by instructors. Multidisciplinary Science Journal, 8(2), 2026021.
https://doi.org/10.31893/multiscience.2026021
Ebadi, S., Velayati, S., Ramezanzadeh, A., & Salman, A. R. (2025). Exploring the impact of AI-
powered speaking tasks on EFL learners' speaking performance and anxiety: An activity
theory study. Acta psychologica, 259, 105391.
https://doi.org/10.1016/j.actpsy.2025.105391
REFERENCES
Alenezi, A., & Alenezi, A. (2025). Evaluating the Effectiveness of Chatbot-Assisted Learning in
Enhancing English Conversational Skills Among Secondary School Students. Education
Sciences, 15(9), 1136. https://doi.org/10.3390/educsci15091136
Aziz, A. A., & Kashinathan, S. (2021). ESL learners’ challenges in speaking English in Malaysian
classroom. Development, 10(2), 983-991. https://doi.org/10.6007/ijarped/v10-i2/10355
Brown, H. D., & Abeywickrama, P. (2019). Language assessment: Principles and classroom
practices (3rd ed.). Pearson.
Bustamante Salinas, P. M., Rodríguez Maida, A. A., & Quisbert Pinedo, L. (2026). Integración
de recursos de inteligencia artificial en los procesos educativos. Revista Tecnología,
Ciencia Y Educación, (33), 139–167. https://doi.org/10.51302/tce.2026.24177
Cha, J., Han, J., Yoo, H., & Oh, A. (2024). CHOP: Integrating ChatGPT into EFL Oral
Presentation Practice. arXiv preprint arXiv:2407.07393.
https://doi.org/10.48550/arXiv.2407.07393
Cohen, L., Manion, L., & Morrison, K. (2018). Research methods in education (8th ed.).
Routledge.
Council of Europe. (2020). Common European Framework of Reference for Languages:
Learning, teaching, assessment—Companion volume. Council of Europe Publishing.
https://www.coe.int/en/web/common-european-framework-reference-languages
Creswell, J. W., & Creswell, J. D. (2018). Research design: Qualitative, quantitative, and mixed
methods approaches (5th ed.). SAGE Publications.
Cuevas-Montero, R. (2021). Desarrollo e integración de las destrezas orales en la educación a
distancia a través de podcasts colaborativos. En J. Martín Párraga & J. M. Garrido Anguita
(Eds.), De Platón al “Homo Tecnologicus”: Las humanidades en el siglo XXI (pp. 75–82).
Tirant Humanidades.
Deng, L., & Jamaludin, K. A. (2026). Roles of Generative Artificial Intelligence (GENAI) in
English as a Foreign Language (EFL) instruction: A systematic literature review. SAGE
Open, 16(1). https://doi.org/10.1177/21582440261418315
Du, T. T., Duc, L. V., The, N. H., & Dung, D. T. B. (2025). Challenges faced by EFL students at
a university in Vietnam in developing English speaking skills and pedagogical strategies
used by instructors. Multidisciplinary Science Journal, 8(2), 2026021.
https://doi.org/10.31893/multiscience.2026021
Ebadi, S., Velayati, S., Ramezanzadeh, A., & Salman, A. R. (2025). Exploring the impact of AI-
powered speaking tasks on EFL learners' speaking performance and anxiety: An activity
theory study. Acta psychologica, 259, 105391.
https://doi.org/10.1016/j.actpsy.2025.105391
Vol. 13/ Núm. 3 2026 pág. 2009
Ericsson, E., & Johansson, S. (2023). English speaking practice with conversational AI: Lower
secondary students’ educational experiences over time. Computers and Education
Artificial Intelligence, 5, 100164. https://doi.org/10.1016/j.caeai.2023.100164
Fathi, J., Rahimi, M., & Derakhshan, A. (2024). Improving EFL learners’ speaking skills and
willingness to communicate via artificial intelligence-mediated interactions. System, 121,
103254. https://doi.org/10.1016/j.system.2024.103254
Fauzi, I., Hartono, R., Rukmini, D., & Pratama, H. (2025). AI Applications for EFL Learners:
Enhancing Speaking Performance and Reducing Anxiety with Gender-Based Analysis.
Forum for Linguistic Studies, 7(9). https://doi.org/10.30564/fls.v7i9.10192
Fernández-García, A., & Fonseca-Mora, M. C. (2019). EFL learners’ speaking proficiency and
its connection to emotional understanding, willingness to communicate and musical
experience. Language Teaching Research, 26(1), 124–140.
https://doi.org/10.1177/1362168819891868
Guerrero, F. S. C., Guillén, L. A. P., Tenesaca, J. R. B., & Herrera, D. C. E. (2026). The Use of
Artificial Intelligence to Improve Speaking Fluency in English as a Foreign Language
(EFL) Learning. Arandu UTIC. https://doi.org/10.69639/arandu.v13i1.1950
Huang, J. (2024). Enhancing EFL speaking feedback with ChatGPT's voice prompts.
International Journal of TESOL Studies, 6(3). https://doi.org/10.58304/ijts.20240302
Huang, Y., & Kakham, P. (2026). The impacts of AI conversational agents on EFL learners' oral
proficiency and foreign language speaking anxiety. Frontiers in
Education. https://doi.org/10.3389/feduc.2026.1799269
Huu, N., Du, T. T., & Hoang, T. T. (2026). Exploring the Impact of Artificial Intelligence on
English Majors’ Speaking Skills: A Case Study at a Vietnamese University. Arab World
English Journal, 3, 166–179. https://doi.org/10.24093/awej/ai3.11
Jeon, J. (2024). Exploring AI chatbot affordances in the EFL classroom: Young learners’
experiences and perspectives. Computer Assisted Language Learning, 37(1–2), 1–26.
https://doi.org/10.1080/09588221.2021.2021241
Kallio, H., Pietilä, A. M., Johnson, M., & Kangasniemi, M. (2016). Systematic methodological
review: Developing a framework for a qualitative semi-structured interview guide. Journal
of Advanced Nursing, 72(12), 2954–2965. https://doi.org/10.1111/jan.13031
Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U.,
Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel,
C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., . . . Kasneci, G. (2023).
ChatGPT for good? On opportunities and challenges of large language models for
education. Learning and Individual Differences, 103, 102274.
https://doi.org/10.1016/j.lindif.2023.102274
Ericsson, E., & Johansson, S. (2023). English speaking practice with conversational AI: Lower
secondary students’ educational experiences over time. Computers and Education
Artificial Intelligence, 5, 100164. https://doi.org/10.1016/j.caeai.2023.100164
Fathi, J., Rahimi, M., & Derakhshan, A. (2024). Improving EFL learners’ speaking skills and
willingness to communicate via artificial intelligence-mediated interactions. System, 121,
103254. https://doi.org/10.1016/j.system.2024.103254
Fauzi, I., Hartono, R., Rukmini, D., & Pratama, H. (2025). AI Applications for EFL Learners:
Enhancing Speaking Performance and Reducing Anxiety with Gender-Based Analysis.
Forum for Linguistic Studies, 7(9). https://doi.org/10.30564/fls.v7i9.10192
Fernández-García, A., & Fonseca-Mora, M. C. (2019). EFL learners’ speaking proficiency and
its connection to emotional understanding, willingness to communicate and musical
experience. Language Teaching Research, 26(1), 124–140.
https://doi.org/10.1177/1362168819891868
Guerrero, F. S. C., Guillén, L. A. P., Tenesaca, J. R. B., & Herrera, D. C. E. (2026). The Use of
Artificial Intelligence to Improve Speaking Fluency in English as a Foreign Language
(EFL) Learning. Arandu UTIC. https://doi.org/10.69639/arandu.v13i1.1950
Huang, J. (2024). Enhancing EFL speaking feedback with ChatGPT's voice prompts.
International Journal of TESOL Studies, 6(3). https://doi.org/10.58304/ijts.20240302
Huang, Y., & Kakham, P. (2026). The impacts of AI conversational agents on EFL learners' oral
proficiency and foreign language speaking anxiety. Frontiers in
Education. https://doi.org/10.3389/feduc.2026.1799269
Huu, N., Du, T. T., & Hoang, T. T. (2026). Exploring the Impact of Artificial Intelligence on
English Majors’ Speaking Skills: A Case Study at a Vietnamese University. Arab World
English Journal, 3, 166–179. https://doi.org/10.24093/awej/ai3.11
Jeon, J. (2024). Exploring AI chatbot affordances in the EFL classroom: Young learners’
experiences and perspectives. Computer Assisted Language Learning, 37(1–2), 1–26.
https://doi.org/10.1080/09588221.2021.2021241
Kallio, H., Pietilä, A. M., Johnson, M., & Kangasniemi, M. (2016). Systematic methodological
review: Developing a framework for a qualitative semi-structured interview guide. Journal
of Advanced Nursing, 72(12), 2954–2965. https://doi.org/10.1111/jan.13031
Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U.,
Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel,
C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., . . . Kasneci, G. (2023).
ChatGPT for good? On opportunities and challenges of large language models for
education. Learning and Individual Differences, 103, 102274.
https://doi.org/10.1016/j.lindif.2023.102274
Vol. 13/ Núm. 3 2026 pág. 2010
Kim, S., & Park, S. (2023). Young Korean EFL Learners’ Perception of Role-Playing Scripts:
ChatGPT vs. Textbooks. Korean Journal of English Language and Linguistics, 23, 1136–
1153. https://doi.org/10.15738/kjell.23..202312.1136
Klímová, B., & Seraj, P. M. I. (2023). The use of chatbots in university EFL settings: Research
trends and pedagogical implications. Frontiers in Psychology, 14, 1131506.
https://doi.org/10.3389/fpsyg.2023.1131506
Koç, F. Ş., & Savaş, P. (2024). The use of artificially intelligent chatbots in English language
learning: A systematic meta-synthesis study of articles published between 2010 and 2024.
ReCALL, 37(1), 4–21. https://doi.org/10.1017/s0958344024000168
Krashen, S. D. (1982). Principles and practice in second language acquisition. Pergamon Press.
Lim, W. M., Gunasekara, A., Pallant, J. L., Pallant, J. I., & Pechenkina, E. (2023). Generative AI
and the future of education: Ragnarök or reformation? A paradoxical perspective from
management educators. The International Journal of Management Education, 21(2),
100790. https://doi.org/10.1016/j.ijme.2023.100790
Lo, C. K., Yu, P., Xu, S., Ng, D., & Jong, M. S. (2024). Exploring the application of ChatGPT in
ESL/EFL education and related research issues: A systematic review of empirical studies.
Smart Learning Environments, 11, Article 49. https://doi.org/10.1186/s40561-024-00342-
5
Luoma, S. (2004). Assessing speaking. Cambridge University Press.
Mageira, K., Pittou, D., Papasalouros, A., Kotis, K., Zangogianni, P., & Daradoumis, A. (2022).
Educational AI chatbots for content and language integrated learning. Applied Sciences,
12(7), 3239. https://doi.org/10.3390/app12073239
McNamara, T. F. (1996). Measuring second language performance. Longman.
Nguyen Thi Minh Nguyet, & Pham Thi Thanh Nhan. (2024). Voice mode of ChatGPT: A new
support tool for teaching English in university programs. EPRA International Journal of
Research & Development (IJRD), 9(7).
https://eprajournals.com/IJSR/article/13628/abstract
Nhan, L. K., Hoa, N. T. M., & Quang, L. V. N. (2025). Revolutionizing speaking skills
improvement: AI’s role in personalized language learning. International Journal of
Innovative Research and Scientific Studies, 8(2), 4637–4649.
https://doi.org/10.53894/ijirss.v8i2.6408
Niglas, K., Teddlie, C., & Tashakkori, A. (2021). The use of mixed methods in educational
research: A review. Quality & Quantity, 55(3), 1007–1025.
https://doi.org/10.1007/s11135-021-01218-3
Núñez, A. A. C., Nuñez, M. S. C., Pachay, J. F. Z. P. Z., & Bosquez, A. M. C. B. C. (2025). Using
ChatGPT Voice to improve speaking skills in English language learners. Ciencia Latina
Kim, S., & Park, S. (2023). Young Korean EFL Learners’ Perception of Role-Playing Scripts:
ChatGPT vs. Textbooks. Korean Journal of English Language and Linguistics, 23, 1136–
1153. https://doi.org/10.15738/kjell.23..202312.1136
Klímová, B., & Seraj, P. M. I. (2023). The use of chatbots in university EFL settings: Research
trends and pedagogical implications. Frontiers in Psychology, 14, 1131506.
https://doi.org/10.3389/fpsyg.2023.1131506
Koç, F. Ş., & Savaş, P. (2024). The use of artificially intelligent chatbots in English language
learning: A systematic meta-synthesis study of articles published between 2010 and 2024.
ReCALL, 37(1), 4–21. https://doi.org/10.1017/s0958344024000168
Krashen, S. D. (1982). Principles and practice in second language acquisition. Pergamon Press.
Lim, W. M., Gunasekara, A., Pallant, J. L., Pallant, J. I., & Pechenkina, E. (2023). Generative AI
and the future of education: Ragnarök or reformation? A paradoxical perspective from
management educators. The International Journal of Management Education, 21(2),
100790. https://doi.org/10.1016/j.ijme.2023.100790
Lo, C. K., Yu, P., Xu, S., Ng, D., & Jong, M. S. (2024). Exploring the application of ChatGPT in
ESL/EFL education and related research issues: A systematic review of empirical studies.
Smart Learning Environments, 11, Article 49. https://doi.org/10.1186/s40561-024-00342-
5
Luoma, S. (2004). Assessing speaking. Cambridge University Press.
Mageira, K., Pittou, D., Papasalouros, A., Kotis, K., Zangogianni, P., & Daradoumis, A. (2022).
Educational AI chatbots for content and language integrated learning. Applied Sciences,
12(7), 3239. https://doi.org/10.3390/app12073239
McNamara, T. F. (1996). Measuring second language performance. Longman.
Nguyen Thi Minh Nguyet, & Pham Thi Thanh Nhan. (2024). Voice mode of ChatGPT: A new
support tool for teaching English in university programs. EPRA International Journal of
Research & Development (IJRD), 9(7).
https://eprajournals.com/IJSR/article/13628/abstract
Nhan, L. K., Hoa, N. T. M., & Quang, L. V. N. (2025). Revolutionizing speaking skills
improvement: AI’s role in personalized language learning. International Journal of
Innovative Research and Scientific Studies, 8(2), 4637–4649.
https://doi.org/10.53894/ijirss.v8i2.6408
Niglas, K., Teddlie, C., & Tashakkori, A. (2021). The use of mixed methods in educational
research: A review. Quality & Quantity, 55(3), 1007–1025.
https://doi.org/10.1007/s11135-021-01218-3
Núñez, A. A. C., Nuñez, M. S. C., Pachay, J. F. Z. P. Z., & Bosquez, A. M. C. B. C. (2025). Using
ChatGPT Voice to improve speaking skills in English language learners. Ciencia Latina
Vol. 13/ Núm. 3 2026 pág. 2011
Revista Científica Multidisciplinar, 9(1), 7143–7161.
https://doi.org/10.37811/cl_rcm.v9i1.16390
Park, H. (2023). Application of ChatGPT for an English learning platform. STEM Journal, 24(3),
30–48. https://doi.org/10.16875/stem.2023.24.3.30
Pau, L. C., Yunus, M. M., & Lun, C. W. W. (2026). The impact of popular artificial intelligence
tools on English language teaching and learning: a systematic literature review (2021-
2025). International Journal of Research and Innovation in Social Science, IX(XII), 3804–
3827. https://doi.org/10.47772/ijriss.2025.91200296
Rao, P. S. (2019). The importance of speaking skills in English classrooms. Alford Council of
International English & Literature Journal, 2(2).
https://www.researchgate.net/publication/334283040_THE_IMPORTANCE_OF_SPEA
KING_SKILLS_IN_ENGLISH_CLASSROOMS
Rodriguez, S. E. G., & Baquerizo, A. S. M. (2025). Ecuadorian EFL teachers’ experiences in
fostering students’ English-speaking skills: insights into strategies and challenges in public
and private schools. UNESUM - Ciencias Revista Científica Multidisciplinaria, 9(2), 124–
136. https://doi.org/10.47230/unesum-ciencias.v9.n2.2025.124-136
Sosa-López, G., & Mora, J. (2022). The role of speaking anxiety on L2 English speaking fluency,
accuracy, and complexity. In J. Levis & A. Guskaroska (Eds.), Proceedings of the 12th
Pronunciation in Second Language Learning and Teaching Conference (held June 2021,
Brock University, St. Catharines, ON). https://doi.org/10.31274/psllt.13362
Susoy, Z. (2026). Reducing anxiety and enhancing performance: the impact of AI chatbots versus
human facilitation on EFL speaking assessment outcomes. Frontiers in Psychology, 16,
1745942. https://doi.org/10.3389/fpsyg.2025.1745942
Tai, T., & Chen, H. H. (2024). Improving elementary EFL speaking skills with generative AI
chatbots: Exploring individual and paired interactions. Computers & Education, 220,
105112. https://doi.org/10.1016/j.compedu.2024.105112
Tan, W., & Ismail, H. H. (2025). Malaysian ESL learners’ experiences with AI-driven voice
practice using ChatGPT and traditional role-play. International Journal of Education
Psychology and Counseling. https://doi.org/10.35631/ijepc.1061086
Tan, X., Cheng, G., & Ling, M. H. (2024). Artificial intelligence in teaching and teacher
professional development: A systematic review. Computers and Education Artificial
Intelligence, 8, 100355. https://doi.org/10.1016/j.caeai.2024.100355
Tapia, L., & Vega, M. (2024). The impact of Content Learning Integrated Language (CLIL) on
the speaking skill in the EFL classroom at the secondary level. Religación, 9(40),
e2401220. https://doi.org/10.46652/rgn.v9i40.1220
Revista Científica Multidisciplinar, 9(1), 7143–7161.
https://doi.org/10.37811/cl_rcm.v9i1.16390
Park, H. (2023). Application of ChatGPT for an English learning platform. STEM Journal, 24(3),
30–48. https://doi.org/10.16875/stem.2023.24.3.30
Pau, L. C., Yunus, M. M., & Lun, C. W. W. (2026). The impact of popular artificial intelligence
tools on English language teaching and learning: a systematic literature review (2021-
2025). International Journal of Research and Innovation in Social Science, IX(XII), 3804–
3827. https://doi.org/10.47772/ijriss.2025.91200296
Rao, P. S. (2019). The importance of speaking skills in English classrooms. Alford Council of
International English & Literature Journal, 2(2).
https://www.researchgate.net/publication/334283040_THE_IMPORTANCE_OF_SPEA
KING_SKILLS_IN_ENGLISH_CLASSROOMS
Rodriguez, S. E. G., & Baquerizo, A. S. M. (2025). Ecuadorian EFL teachers’ experiences in
fostering students’ English-speaking skills: insights into strategies and challenges in public
and private schools. UNESUM - Ciencias Revista Científica Multidisciplinaria, 9(2), 124–
136. https://doi.org/10.47230/unesum-ciencias.v9.n2.2025.124-136
Sosa-López, G., & Mora, J. (2022). The role of speaking anxiety on L2 English speaking fluency,
accuracy, and complexity. In J. Levis & A. Guskaroska (Eds.), Proceedings of the 12th
Pronunciation in Second Language Learning and Teaching Conference (held June 2021,
Brock University, St. Catharines, ON). https://doi.org/10.31274/psllt.13362
Susoy, Z. (2026). Reducing anxiety and enhancing performance: the impact of AI chatbots versus
human facilitation on EFL speaking assessment outcomes. Frontiers in Psychology, 16,
1745942. https://doi.org/10.3389/fpsyg.2025.1745942
Tai, T., & Chen, H. H. (2024). Improving elementary EFL speaking skills with generative AI
chatbots: Exploring individual and paired interactions. Computers & Education, 220,
105112. https://doi.org/10.1016/j.compedu.2024.105112
Tan, W., & Ismail, H. H. (2025). Malaysian ESL learners’ experiences with AI-driven voice
practice using ChatGPT and traditional role-play. International Journal of Education
Psychology and Counseling. https://doi.org/10.35631/ijepc.1061086
Tan, X., Cheng, G., & Ling, M. H. (2024). Artificial intelligence in teaching and teacher
professional development: A systematic review. Computers and Education Artificial
Intelligence, 8, 100355. https://doi.org/10.1016/j.caeai.2024.100355
Tapia, L., & Vega, M. (2024). The impact of Content Learning Integrated Language (CLIL) on
the speaking skill in the EFL classroom at the secondary level. Religación, 9(40),
e2401220. https://doi.org/10.46652/rgn.v9i40.1220
Vol. 13/ Núm. 3 2026 pág. 2012
Wei, L. (2023). Artificial intelligence in language instruction: impact on English learning
achievement, L2 motivation, and self-regulated learning. Frontiers in Psychology, 14,
1261955. https://doi.org/10.3389/fpsyg.2023.1261955
Wei, X. (2022). Mixed methods research in language education: Applications and trends.
Language Education and Research, 5(2), 14–23. https://ojs.piscomed.com/index.php/L-
E/article/view/3103
Xing, C. (2025). A Systematic Review on Artificial intelligence (AI) Technologies in ESL/EFL
speaking Skills. International Journal of TESOL Studies.
https://doi.org/10.58304/ijts.250908
Yan, H., & Singh, M. K. S. (2026). The impact of AI-mediated instruction on speaking
proficiency, enjoyment, anxiety, and emotional engagement: A mixed-methods approach.
Humanities & Social Sciences Communications. https://doi.org/10.1057/s41599-026-
06705-2
Zhai, C., & Wibowo, S. (2023). A systematic review on artificial intelligence dialogue systems
for enhancing English as foreign language students’ interactional competence in the
university. Computers and Education Artificial Intelligence, 4, 100134.
https://doi.org/10.1016/j.caeai.2023.100134
Zhang, C., Meng, Y., & Ma, X. (2024). Artificial intelligence in EFL speaking: Impact on
enjoyment, anxiety, and willingness to communicate. System, 121, 103259.
https://doi.org/10.1016/j.system.2024.103259
Shazly, R. E. (2021). Effects of artificial intelligence on English speaking anxiety and speaking
performance: A case study. Expert Systems, 38. https://doi.org/10.1111/exsy.12667
Wei, L. (2023). Artificial intelligence in language instruction: impact on English learning
achievement, L2 motivation, and self-regulated learning. Frontiers in Psychology, 14,
1261955. https://doi.org/10.3389/fpsyg.2023.1261955
Wei, X. (2022). Mixed methods research in language education: Applications and trends.
Language Education and Research, 5(2), 14–23. https://ojs.piscomed.com/index.php/L-
E/article/view/3103
Xing, C. (2025). A Systematic Review on Artificial intelligence (AI) Technologies in ESL/EFL
speaking Skills. International Journal of TESOL Studies.
https://doi.org/10.58304/ijts.250908
Yan, H., & Singh, M. K. S. (2026). The impact of AI-mediated instruction on speaking
proficiency, enjoyment, anxiety, and emotional engagement: A mixed-methods approach.
Humanities & Social Sciences Communications. https://doi.org/10.1057/s41599-026-
06705-2
Zhai, C., & Wibowo, S. (2023). A systematic review on artificial intelligence dialogue systems
for enhancing English as foreign language students’ interactional competence in the
university. Computers and Education Artificial Intelligence, 4, 100134.
https://doi.org/10.1016/j.caeai.2023.100134
Zhang, C., Meng, Y., & Ma, X. (2024). Artificial intelligence in EFL speaking: Impact on
enjoyment, anxiety, and willingness to communicate. System, 121, 103259.
https://doi.org/10.1016/j.system.2024.103259
Shazly, R. E. (2021). Effects of artificial intelligence on English speaking anxiety and speaking
performance: A case study. Expert Systems, 38. https://doi.org/10.1111/exsy.12667