Tipo de artículo: Artículo original

Fecha de recibido: 26/05/2026

Fecha de aceptado: 28/08/2026

Fecha de publicado: 01/09/2026

Introduction

Written conversation in English is an essential competence for university students in teacher education programs because it combines linguistic accuracy, communicative intention, coherence, and confidence in interaction. In English as a Foreign Language contexts, students often know vocabulary and grammar rules but still find it difficult to maintain written exchanges fluently. This difficulty is associated with limited practice outside the classroom, delayed feedback, fear of making mistakes, and uncertainty about how to improve phrases, expressions, grammar, and vocabulary during authentic interaction.

The accelerated incorporation of artificial intelligence in education has opened new possibilities for language learning. AI-based tools such as conversational chatbots, automated correctors, generative language models, translators, and writing assistants can provide immediate responses, adaptive suggestions, and opportunities for autonomous practice. In language education, recent reviews have identified a growing research interest in the role of AI as tutor, assessment tool, feedback provider, and learning partner (Liang et al., 2023); (Al-khresheh, 2024).

However, the educational value of AI cannot be assumed merely because the technology is available. Large language models and chatbots offer useful opportunities for feedback, examples, reformulation, and conversational practice, but they also present risks related to misinformation, overdependence, superficial revision, academic integrity, and limited human interaction. Therefore, their integration in higher education requires critical guidance, digital literacy, and clear pedagogical purposes (Kasneci et al., 2023); (Miao & Holmes, 2023).

In the Pedagogy of National and Foreign Languages program, future English teachers need to develop written communicative competence while learning how to use technological resources responsibly. For this reason, it is relevant to examine how students perceive AI tools as support for written English conversation practice. The study does not attempt to prove the causal effectiveness of AI on writing performance; instead, it describes students’ perceptions of usefulness, acceptance, benefits, and limitations in a specific university context.

The research question guiding the study was: How do fifth-semester students perceive the use of artificial intelligence tools as support for written English conversation practice? Accordingly, the objective was to describe students’ perceptions of AI tools as complementary resources for practicing written English conversation in a higher education EFL context. The contribution of the study lies in providing contextual evidence for language teacher education, where AI is increasingly present but still requires pedagogical mediation.

Theoretical and conceptual basis

Artificial intelligence in language education refers to the use of computational systems capable of processing linguistic input, generating responses, providing feedback, recommending resources, and adapting tasks to learner performance. In EFL contexts, AI can support writing and vocabulary development by offering examples, paraphrases, grammar explanations, corrective suggestions, and interactive practice. Nevertheless, AI should be understood as a pedagogical resource rather than as an autonomous substitute for the teacher.

Generative AI tools, especially chatbots, are relevant for written English conversation because they can simulate exchanges, maintain topic continuity, and respond to learner input in real time. ChatGPT and similar tools have been discussed as language teaching resources because they can generate model dialogues, reformulate sentences, explain vocabulary, and provide additional practice opportunities. Their usefulness depends on learners’ ability to formulate prompts, evaluate responses, and transform suggestions into learning (Kohnke et al., 2023).

Feedback is one of the most important pedagogical dimensions of AI-supported writing practice. Automated systems can provide individualized comments and adaptive suggestions, which may help students revise texts and notice errors. Meta-analytic evidence on automated writing evaluation suggests positive effects on writing quality, although the interpretation of feedback and the quality of revision remain central to learning (Zhai & Ma, 2023); (Fleckenstein et al., 2023).

Studies focused on AI-generated feedback have also shown that students may value immediate comments, alternative wordings, and explanations during writing tasks. However, AI feedback should not be treated as automatically accurate or pedagogically sufficient. Students need teacher guidance to understand which suggestions are appropriate, how to maintain their authorial voice, and how to avoid uncritical dependence on automated correction (Escalante et al., 2023); (Mahapatra, 2024).

AI tools can also support vocabulary development and phrase construction. Technology-assisted vocabulary learning has shown positive effects in L2 contexts, especially when learners engage with multimodal resources and meaningful practice. AI-based applications can provide contextualized vocabulary, object-based translation, pronunciation support, and sentence-level examples, which may help students move beyond isolated word memorization (Yu & Trainin, 2022); (Liu & Chen, 2023).

Despite these affordances, responsible use remains essential. AI may generate inaccurate information, oversimplified expressions, culturally inappropriate examples, or feedback that does not match the communicative purpose of a text. For future English teachers, this means AI should be integrated with reflective tasks, ethical guidance, and opportunities to compare automated feedback with teacher and peer feedback.

Materials and methods

This study was conducted under a quantitative approach with a basic descriptive scope. The approach was quantitative because the data were collected through closed-ended questions and processed using frequencies and percentages. The scope was descriptive because the purpose was to characterize students’ perceptions and practices regarding the use of artificial intelligence tools for written English conversation practice, without testing an experimental intervention or establishing causal relationships.

The design was non-experimental and cross-sectional. It was non-experimental because no variables were manipulated, no treatment group was created, and no controlled pedagogical intervention was applied. The phenomenon was observed in its natural academic context. It was cross-sectional because the information was collected at a single point during the academic period.

The study was carried out at Universidad Estatal del Sur de Manabí, in the Pedagogy of National and Foreign Languages program. The participants were 27 fifth-semester students from group B, morning schedule. Since the survey was applied to the available intact classroom group, the study used a census of that group. This methodological decision was coherent with the descriptive purpose of the research and limits interpretation to this specific educational context.

The data collection technique was a survey. The instrument was a structured questionnaire composed of closed-ended items organized around the following dimensions: confidence and reduction of fear when writing in English, improvement of phrases and expressions, autonomous learning and conversation practice, conversational adaptation, motivation to write, commitment to practice, grammatical and spelling accuracy, and vocabulary expansion. The response scale included four levels of perceived importance: not at all important, slightly important, moderately important, and very important.

The research procedure followed four stages. First, the problem and objective were defined based on the classroom project and the need to describe students’ perceptions of AI-supported written conversation practice. Second, the questionnaire items were organized according to the dimensions of the study. Third, the instrument was applied digitally to the 27 participating students. Fourth, the responses were tabulated and interpreted through descriptive statistics.

Frequencies and percentages were calculated for each item. No inferential statistical tests were applied because the objective was not predictive, comparative, or causal. Consequently, the findings should be interpreted as contextual evidence of students’ perceptions rather than as proof of improvement in written English performance. Ethical considerations included voluntary participation, academic use of the information, and non-identification of individual responses in the presentation of results.

Results and discussion

The results are presented in two integrated tables to comply with the journal guideline that limits the number of tables, figures, or charts. The first table groups perceptions related to confidence, phrase improvement, autonomous practice, and conversation adaptation. The second table groups perceptions related to motivation, commitment, accuracy, and vocabulary expansion.

Table 1. Students’ perceptions of AI support for confidence, expression improvement, autonomy and conversation adaptation

Item	Not at all important	Slightly important	Moderately important	Very important
Confidence when writing and reduction of fear of mistakes	1 (4%)	2 (7%)	18 (67%)	6 (22%)
Suggestions to improve English phrases and expressions	2 (7%)	2 (7%)	17 (63%)	6 (22%)
Autonomous learning and conversation practice outside class	1 (4%)	3 (11%)	15 (56%)	8 (30%)
Adaptation of written conversation to learner needs	1 (4%)	3 (11%)	12 (44%)	11 (41%)

Note. Data from the survey applied to 27 fifth-semester students from group B of the PINE program.

Table 1 shows that most students perceived AI tools as moderately or very important in the four analyzed dimensions. The highest concentration appeared in confidence and reduction of fear of making mistakes, where 67% selected moderately important and 22% selected very important. This finding suggests that students associate AI with a low-pressure practice environment where they can test phrases, receive suggestions, and reduce anxiety before writing in English.

The second item also showed a positive tendency: 63% considered AI suggestions for improving phrases and expressions moderately important and 22% considered them very important. This result is coherent with research on AI-generated writing feedback, which indicates that students often value immediate reformulations, explanations, and alternatives when revising texts. (Escalante et al., 2023); (Mahapatra, 2024)

Autonomous learning was also valued positively, with 56% selecting moderately important and 30% selecting very important. This result indicates that students perceived AI as a resource for practice beyond classroom time. Nevertheless, autonomy should not be confused with unguided dependence on tools. Learners need criteria to decide when AI responses are accurate, useful, and appropriate for the communicative purpose of the text (Miao & Holmes, 2023).

Conversational adaptation obtained a notably high positive perception: 44% selected moderately important and 41% selected very important. This finding suggests that students recognized the value of AI tools for adjusting responses, maintaining context, and providing flexible written interaction. Research on ChatGPT for language teaching has similarly highlighted the potential of generative AI to support contextualized practice, while also warning that teachers and learners need digital competencies to use it effectively (Kohnke et al., 2023).

Table 2. Students’ perceptions of AI support for motivation, accuracy and vocabulary expansion.

Item	Not at all important	Slightly important	Moderately important	Very important
Motivation to practice written English conversation	0 (0%)	3 (11%)	7 (26%)	17 (63%)
Commitment and motivation in English practice	1 (4%)	2 (7%)	14 (52%)	10 (37%)
Grammatical and spelling accuracy	0 (0%)	4 (15%)	12 (44%)	11 (41%)
Vocabulary expansion and language learning	0 (0%)	5 (19%)	10 (37%)	12 (44%)

Note. Data from the survey applied to 27 fifth-semester students from group B of the PINE program.

Table 2 presents an even stronger tendency in the motivational dimension. Sixty-three percent of students considered AI very important for motivating them to practice written English conversation, while 26% selected moderately important. This indicates that AI tools may be attractive for learners because they provide immediacy, availability, and interactive practice opportunities. Similar studies have reported favorable student perceptions of AI applications in writing classes, although they also emphasize the importance of careful pedagogical integration. (Sumakul et al., 2022)

Commitment and motivation also received positive responses: 52% selected moderately important and 37% selected very important. This result suggests that students perceived AI as a support for sustained practice rather than as an occasional tool. From the perspective of teacher education, this is relevant because future English teachers need to experience digital tools as pedagogical resources, not only as users seeking quick corrections.

Regarding grammatical and spelling accuracy, 44% selected moderately important and 41% selected very important. This finding is consistent with research on automated writing evaluation and AI-powered writing assistants, which indicates that automated feedback can help learners identify errors and revise written products. However, the result should be interpreted cautiously because the present study did not evaluate actual writing samples. (Nazari et al., 2021); (Zhai & Ma, 2023)

Vocabulary expansion was also positively perceived, with 37% selecting moderately important and 44% selecting very important. This finding is consistent with studies on technology-assisted vocabulary learning and AI-based vocabulary applications, which suggest that digital tools can provide contextualized input, multimodal support, and repeated exposure to lexical items. (Yu & Trainin, 2022); (Liu & Chen, 2023)

Taken together, the results indicate a favorable perception of AI as a complementary resource for written English conversation practice. Nevertheless, the findings do not demonstrate that AI caused measurable improvement in writing performance. They show that students perceived AI as useful for confidence, motivation, grammar, vocabulary, and autonomous practice. This distinction is important because perception-based descriptive studies cannot replace experimental or performance-based evidence.

The findings also point to the need for responsible pedagogical integration. If students use AI only to obtain corrected sentences, they may become dependent on automatic suggestions. If they use AI as a reflective tool, they can compare alternatives, ask for explanations, identify patterns in their errors, and revise intentionally. Therefore, teacher mediation is essential to transform AI interaction into language learning.

For language teacher education, the implications are clear. Future English teachers should learn how to design AI-supported writing activities, how to evaluate AI-generated feedback, and how to promote academic integrity. Classroom use of AI should include prompt design, revision logs, comparison between AI feedback and teacher feedback, and reflection on why specific suggestions were accepted or rejected.

The main limitation of this study is its contextual scope. The participants were 27 students from one intact classroom group; therefore, the results cannot be generalized to all students in the program or to other institutions. Another limitation is the use of self-reported data. The questionnaire describes perceptions but does not directly measure writing quality, fluency, lexical development, or grammatical improvement. Future studies should include writing samples, analytic rubrics, pretest and posttest designs, and larger samples.

Pedagogical implications for AI-supported EFL writing

The first pedagogical implication is that AI-supported writing tasks should be designed as processes rather than as isolated correction activities. Students should be asked to draft, interact with AI, evaluate the feedback, revise, and explain the decisions made in their final version. This sequence can help learners move from passive acceptance of suggestions to conscious language analysis. (Escalante et al., 2023)

A second implication concerns prompt literacy. Students do not necessarily know how to request useful feedback, ask for clarification, or delimit the level and purpose of an AI-generated response. Teacher education programs should therefore include activities in which future teachers compare different prompts and evaluate the quality of the language support produced by AI tools. (Kohnke et al., 2023)

A third implication is related to formative assessment. AI tools can provide rapid feedback, but classroom assessment should still include teacher criteria, rubrics, and reflective evidence. When feedback is combined with rubrics and revision logs, learners can better understand not only what should be corrected but also why a change improves the communicative quality of the text. (Fleckenstein et al., 2023); (Zhai & Ma, 2023)

A fourth implication concerns academic integrity and responsible use. Students should be guided to distinguish between using AI for language support and delegating authorship. In written conversation practice, AI can serve as a partner for rehearsal, vocabulary exploration, and phrase improvement, but the learner must remain responsible for the final communicative decisions. (Miao & Holmes, 2023); (Kasneci et al., 2023)

Finally, the results suggest that AI integration in EFL writing should be accompanied by human interaction. The goal is not to replace teachers, peers, or authentic communication, but to increase opportunities for rehearsal and revision. This balance is particularly important in teacher education because future educators need to understand both the affordances and the limitations of AI-mediated language practice. (Al-khresheh, 2024)

Recommendations for future research

Future research should incorporate direct evidence of writing performance. A useful design would include an initial writing task, a guided AI-supported practice sequence, and a final writing task evaluated with an analytic rubric. This would allow researchers to examine changes in coherence, vocabulary use, grammatical accuracy, and interactional appropriateness. (Mahapatra, 2024)

It would also be valuable to compare different AI tools. Chatbots, grammar checkers, translation tools, and automated writing evaluation systems do not offer the same type of feedback. A comparative design could identify which tools are more useful for phrase reformulation, vocabulary expansion, conversation simulation, or revision accuracy. (Liang et al., 2023); (Liu & Chen, 2023)

Another possible line of research is the analysis of students’ prompts and revision decisions. Instead of focusing only on final texts, researchers could examine how students interact with AI, what kinds of questions they ask, which suggestions they accept, and how they justify their revisions. This would provide a deeper understanding of AI-supported writing as a learning process. (Kohnke et al., 2023)

Finally, future studies should consider mixed-method approaches. Quantitative surveys can describe general tendencies, but interviews, focus groups, think-aloud protocols, and writing portfolios can explain why students trust, question, or reject AI-generated suggestions. This would strengthen the interpretation of findings and contribute to a more balanced understanding of AI in language learning. (Sumakul et al., 2022)

Conclusions

Students perceived artificial intelligence tools as useful complementary resources for written English conversation practice, especially because they provided immediate suggestions, reduced fear of making mistakes, and supported autonomous practice.

The strongest perceptions were related to motivation, conversational adaptation, grammatical and spelling accuracy, and vocabulary expansion. These results suggest that AI tools can encourage students to practice English writing outside the classroom.

The findings should be interpreted as descriptive evidence from one intact classroom group, not as proof that AI improved students’ writing performance. The study describes perceived usefulness rather than causal effectiveness.

Teacher guidance remains necessary. AI tools can support written conversation practice, but they should be integrated with pedagogical criteria, ethical use, reflective revision, and human feedback.

Future research should expand the sample, include direct assessment of written products, compare AI-supported and non-AI-supported writing practice, and analyze how students transform AI-generated suggestions into actual language learning.

Referencias

Al-khresheh, M. H. (2024). Bridging technology and pedagogy from a global lens: Teachers’ perspectives on integrating ChatGPT in English language teaching. Computers and Education: Artificial Intelligence, 6, 100218. https://doi.org/10.1016/j.caeai.2024.100218

Escalante, J., Pack, A., & Barrett, A. (2023). AI-generated feedback on writing: Insights into efficacy and ENL student preference. International Journal of Educational Technology in Higher Education, 20, 57. https://doi.org/10.1186/s41239-023-00425-2

Fleckenstein, J., Liebenow, L. W., & Meyer, J. (2023). Automated feedback and writing: A multi-level meta-analysis of effects on students’ performance. Frontiers in Artificial Intelligence, 6, 1162454. https://doi.org/10.3389/frai.2023.1162454

Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., & Fischer, F. (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

Kohnke, L., Moorhouse, B. L., & Zou, D. (2023). ChatGPT for language teaching and learning. RELC Journal, 54(2), 537-550. https://doi.org/10.1177/00336882231162868

Liang, J. C., Hwang, G. J., Chen, M. R. A., & Darmawansah, D. (2023). Roles and research foci of artificial intelligence in language education: An integrated bibliographic analysis and systematic review approach. Interactive Learning Environments, 31(7), 4270-4296. https://doi.org/10.1080/10494820.2021.1958348

Liu, P. L., & Chen, C. J. (2023). Using an AI-based object detection translation application for English vocabulary learning. Educational Technology & Society, 26(3), 5-20. https://doi.org/10.30191/ETS.202307_26(3).0002

Mahapatra, S. (2024). Impact of ChatGPT on ESL students’ academic writing skills: A mixed methods intervention study. Smart Learning Environments, 11, 9. https://doi.org/10.1186/s40561-024-00295-9

Miao, F., & Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research

Nazari, N., Shabbir, M. S., & Setiawan, R. (2021). Application of artificial intelligence powered digital writing assistant in higher education: Randomized controlled trial. Heliyon, 7(5), e07014. https://doi.org/10.1016/j.heliyon.2021.e07014

Sumakul, D. T. Y. G., Hamied, F. A., & Sukyadi, D. (2022). Students’ perceptions of the use of AI in a writing class. In Proceedings of the 67th TEFLIN International Virtual Conference & the 9th ICOELT 2021 (pp. 52-57). https://doi.org/10.2991/assehr.k.220201.009

Yu, A., & Trainin, G. (2022). A meta-analysis examining technology-assisted L2 vocabulary learning. ReCALL, 34(2), 235-252. https://doi.org/10.1017/S0958344021000239

Zhai, N., & Ma, X. (2023). The effectiveness of automated writing evaluation on writing quality: A meta-analysis. Journal of Educational Computing Research, 61(4), 875-900. https://doi.org/10.1177/07356331221127300