Bridging the Gap: The Role of AI-Powered Pronunciation Assistants in EFL Learning

1. Introduction

Pronunciation remains a fundamental yet frequently underdeveloped component of English as a Foreign Language (EFL) learning. Although communicative competence requires intelligible speech, pronunciation instruction may receive less attention than grammar, vocabulary, and writing. Limited opportunities for meaningful oral interaction can further restrict learners’ pronunciation practice. Recent advances in artificial intelligence (AI), particularly automatic speech recognition (ASR) and automated speech evaluation, offer opportunities for repeated speaking practice and immediate feedback. Zou et al. (2023) demonstrated the potential of AI based speech evaluation to support EFL speaking development, while Dennis (2024) examined AI powered speech recognition for pronunciation instruction. However, access to these tools does not guarantee meaningful learning. Their value depends on how learners interpret feedback and incorporate it into their practice (Muthmainnah et al., 2025).

This issue is particularly relevant in Indonesia, where English generally functions as a foreign language and opportunities for sustained English interaction vary across educational settings. Learners may depend heavily on classroom instruction, while teachers must address diverse learning needs within limited instructional time. AI powered pronunciation assistants can extend practice beyond the classroom through flexible speaking activities and automated feedback. Nevertheless, recognition errors and unclear suggestions may complicate learners’ decisions about pronunciation improvement. UNESCO (2023) emphasizes that AI integration in education should remain human centred, prioritizing pedagogical purpose, equity, and learner agency. Research in Indonesian EFL contexts should therefore examine how learners engage with these tools and the challenges they encounter.

Pedagogically, AI powered pronunciation assistants may help bridge the gap between learners’ need for repeated practice and the limited individual feedback available during lessons. Previous studies have investigated automated speech evaluation and AI supported pronunciation development (Alek, Marzuki, Hidayat, & Sari, 2020; Dennis, 2024; Zou et al., 2023). Research on AI mediated speaking activities has also considered learners’ experiences and willingness to communicate (Muthmainnah, Cardoso et al., 2024; Santiana & Marzuki, 2024; Erizar et al., 2024). However, pronunciation learning involves more than accepting system scores or repeating utterances until they are recognised. Learners must judge whether feedback is appropriate and determine which aspects of their speech require revision. Excessive attention to recognition scores may also overlook stress, rhythm, intonation, and communicative context. Understanding how learners interpret and apply automated feedback is therefore essential.

Despite growing research interest, questions remain about how learners experience and respond to automated pronunciation feedback. Liu et al. (2025), in a systematic review of 24 empirical studies, identified a predominance of quasi experimental designs and a focus on pronunciation accuracy, particularly at the segmental level. They also highlighted limited theoretical grounding and the need for further research into classroom implementation. Fanania et al. (2024) examined students’ perceptions and challenges when using AI based applications for pronunciation practice in Indonesia. Although these studies provide a useful foundation, performance measures and general perceptions cannot fully explain how learners interpret corrections, respond to conflicting feedback, or decide whether to trust an application. Qualitative inquiry can address this gap by examining learners’ interpretations, decisions, and difficulties within their learning contexts.

The present study addresses this gap through a qualitative, process oriented examination of EFL learners’ experiences with AI powered pronunciation assistants. Rather than primarily measuring performance gains, it investigates how learners identify pronunciation difficulties, interpret automated suggestions, evaluate their reliability, and adjust their practice. It also explores tensions between immediate feedback and the limitations of automated recognition. This focus extends existing research by examining the processes underlying learners’ engagement with AI supported pronunciation practice. Consistent with UNESCO’s (2023) human centred perspective, the study treats AI as a supplementary pedagogical resource rather than a replacement for teachers. Its findings may inform instructional decisions, pronunciation task design, and the responsible integration of AI into EFL learning.

Investigating these issues is timely as AI assisted language learning continues to develop. In Indonesian EFL settings, understanding learners’ experiences may help educators extend oral practice while preserving teachers’ roles in contextual explanation and communicative guidance. Accordingly, this study explores learners’ experiences, interpretations, and responses to AI powered pronunciation assistance without presuming that automated feedback necessarily improves pronunciation. It addresses three research questions: (1) How do EFL learners experience AI powered pronunciation assistants in their pronunciation practice? (2) How do learners interpret and respond to feedback generated by these tools? (3) What benefits, challenges, and learning strategies do learners identify when incorporating AI powered pronunciation assistants into EFL learning?

2. Literature Review

2.1. AI-Powered Pronunciation Assistants in EFL Learning

AI powered pronunciation assistants are increasingly relevant to EFL learning because they enable learners to practise spoken English and receive automated feedback beyond classroom instruction. These applications use automatic speech recognition (ASR) and speech analysis to identify aspects of pronunciation that may require attention. Zou et al. (2023) reported improved speaking performance and generally positive perceptions among 40 EFL learners. However, speaking development involves pronunciation, fluency, rhythm, and intelligibility. Numerical scores or sound corrections alone cannot explain how learners interpret feedback and apply it to subsequent speech. Similarly, Liu et al. (2025), reviewing 24 empirical studies, found that ASR research had focused largely on pronunciation accuracy, particularly segmental features. These findings highlight the need to examine AI assistants not only as technological tools but also as learning resources that shape learners’ efforts to develop intelligible and contextually appropriate speech.

2.2. Automated Feedback, Learner Interpretation, and Pronunciation Practice

Feedback is central to pronunciation learning because it helps learners identify differences between their current speech and intended pronunciation. AI powered assistants provide immediate responses, enabling repeated practice and revision without waiting for teacher feedback. However, immediate feedback does not guarantee accurate interpretation or productive use. Zou et al. (2023) found that learners perceived automated speech evaluation as helpful for pronunciation and oral fluency, but positive perceptions alone cannot explain how learners respond to specific corrections. They may accept system judgments, repeat utterances mechanically, or question results when recognition fails. Research on mobile assisted peer feedback among Chinese EFL learners (2022) also highlights the value of examining feedback interactions alongside learning outcomes. These perspectives support investigating how learners interpret AI generated suggestions, assess their relevance, and decide which aspects of their pronunciation to revise.

2.3. Learner Agency, Trust, and the Limitations of AI Feedback

The educational value of AI powered pronunciation assistants depends on learners’ ability to make informed decisions. Learners must decide whether to accept corrections, repeat utterances, seek clarification, or practise independently, particularly when recognition results are inconsistent. Liu et al. (2025) highlighted the limited use of theoretical frameworks and the need to examine how ASR can be integrated into classroom practice. Similarly, UNESCO (2023) emphasizes human agency, pedagogical purpose, and responsible AI use. Learners should therefore evaluate automated feedback critically rather than accept system judgments unconditionally. A qualitative investigation can explore how learners develop trust, manage uncertainty, and use feedback while retaining responsibility for their learning.

2.4. Research Gap and the Positioning of the Present Study

Existing research provides a foundation for investigating AI supported pronunciation learning, but learners’ experiences remain insufficiently understood. Zou et al. (2023) examined speaking outcomes and learner perceptions, while Liu et al. (2025) reviewed ASR research in EFL pronunciation. However, measurable outcomes alone cannot explain how learners interpret feedback, respond to recognition errors, or adapt their practice. This study addresses this gap through a qualitative exploration of learners’ experiences with AI powered pronunciation assistants, focusing on feedback interpretation, trust in automated judgments, and strategies for addressing pronunciation difficulties. Rather than assuming these tools are inherently effective, the study examines their perceived benefits and limitations. Its findings may inform EFL teaching by highlighting how AI feedback can support independent practice while retaining the importance of teacher guidance and meaningful oral communication.

3. Method

3.1. Research Context

This study adopted a qualitative case study design to explore EFL learners’ experiences with AI powered pronunciation assistants. The qualitative approach focused on learners’ interpretations, responses, and learning strategies rather than measuring statistical effectiveness. The study was conducted in an Indonesian EFL context, where learners primarily encountered English through formal instruction and supplementary activities. Previous research on AI speech evaluation and automated feedback provided a foundation for examining both the potential and limitations of these tools (Zou et al., 2023; Amalia & Marzuki, 2023; Amalia et al., 2024; Anita et al., 2024; Apriani et al., 2025). Participants used the selected application to practise individual words, phrases, sentences, and short spoken responses. The researcher examined how learners interpreted automated feedback, responded to unclear or inconsistent recognition results, and adjusted their practice. This contextual approach helped explain how technological assistance interacted with learner agency and the practical challenges of pronunciation learning.

3.2. Participants

The participants were 18 undergraduate students aged 18 to 22 enrolled in an English language education programme at an Indonesian university. They were recruited through purposive sampling based on their involvement in English learning and their experience with, or readiness to use, AI powered pronunciation assistants. Students with different levels of speaking proficiency, pronunciation confidence, and digital familiarity were included to capture diverse perspectives on automated feedback (Marzuki et al., 2018; Muthmainnah et al., 2024; Muthmainnah, Darmawati et al., 2024). The sample size reflected the exploratory nature of the case study, which prioritised the depth and relevance of participants’ experiences over statistical representativeness. Before data collection, the researcher explained the study’s purpose, procedures, voluntary participation, and confidentiality arrangements. Informed consent was obtained, and participants were assured that participation or withdrawal would not affect their academic assessment. Pseudonyms or participant codes were used to protect their identities. This sampling strategy enabled an in depth exploration of shared experiences and differences in how learners interpreted feedback and approached AI assisted pronunciation practice within the selected educational context (Albana et al., 2020; Alek, Marzuki, Farkhan, & Deni, 2020; Alek et al., 2020; Alek et al., 2022).

3.3. Instruments

Three instruments were used: semi structured interviews, non participant observations, and digital artefact analysis. Interviews explored learners’ perceptions of automated feedback, confidence in system judgments, pronunciation difficulties, and strategies for evaluating corrections. Follow up questions addressed inaccurate recognition and unclear feedback (Marzuki, 2019a; Marzuki, 2019b; Santiana et al., 2021; Santiana et al., 2024). Interviews were conducted in Indonesian or English, audio recorded with permission, and transcribed for analysis. Observations documented learners’ interactions with the applications, responses to feedback, and pronunciation adjustments. Digital artefacts, including screenshots and speech recognition outputs, provided further evidence of tool use. Comparing these sources helped contextualise learners’ reported experiences and observed practices (Iftitah et al., 2020; Kuliahana & Marzuki, 2020; Kuliahana et al., 2024; Kuliahana, Marzuki, & Rustam, 2024; Kuliahana & Marzuki, 2024). The instruments were reviewed to ensure alignment with the research questions without assuming that AI powered pronunciation assistants were inherently beneficial or problematic.

3.4. Data Analysis

The data were analysed using reflexive thematic analysis following Braun and Clarke’s (2021) six phases. Interview transcripts, observation notes, and digital artefacts were reviewed repeatedly before initial codes were generated and grouped into preliminary themes. These themes were refined and interpreted in relation to the research questions. The analysis was primarily inductive, supported by reflexive notes and comparisons across data sources. Following Braun and Clarke (2021) and Byrne (2022), the researcher considered contradictory accounts and documented analytical decisions. The final analysis focused on learners’ experiences with AI powered pronunciation assistants, their interpretations of automated feedback, and their perceived benefits, challenges, and learning strategies.

4. Results

4.1. Overview of the Qualitative Findings

The qualitative analysis was designed to examine three aspects of learners’ experiences with AI powered pronunciation assistants: their experiences during pronunciation practice, their interpretations of automated feedback, and the benefits, challenges, and strategies they associated with using these tools. Following Braun and Clarke’s (2021) reflexive thematic analysis, the researcher examined interview transcripts, observation notes, and digital learning artefacts to identify patterns of meaning across the dataset. The preliminary analytical framework presented in this section contains five potential themes: immediate feedback and repeated practice, awareness of pronunciation difficulties, trust and uncertainty in automated recognition, learner agency and independent learning, and the continuing importance of teacher guidance. These themes offer a structure for organizing the findings, but their inclusion as final results requires confirmation against the actual data. The analysis should retain themes that are supported by recurring or analytically significant evidence and should revise or remove those that do not adequately represent participants’ accounts.

Table 1. Illustrative thematic framework for organizing the findings.

Analytical theme	Main analytical focus	Evidence to examine
Immediate feedback and repeated practice	Learners’ responses to instant feedback	Interview accounts, practice observations
Awareness of pronunciation difficulties	Recognition of previously unnoticed problems	Reflections, recorded attempts, interview accounts
Trust and uncertainty	Acceptance or questioning of automated judgments	Feedback screenshots, learner explanations
Learner agency and independent learning	Decisions about revision and further practice	Practice records, observation notes
Teacher guidance and contextual interpretation	Situations requiring human explanation	Interviews, observation notes

Table 1. Illustrative thematic framework for organizing the findings.

Analytical theme	Main analytical focus	Evidence to examine
Immediate feedback and repeated practice	Learners’ responses to instant feedback	Interview accounts, practice observations
Awareness of pronunciation difficulties	Recognition of previously unnoticed problems	Reflections, recorded attempts, interview accounts
Trust and uncertainty	Acceptance or questioning of automated judgments	Feedback screenshots, learner explanations
Learner agency and independent learning	Decisions about revision and further practice	Practice records, observation notes
Teacher guidance and contextual interpretation	Situations requiring human explanation	Interviews, observation notes

4.2. Immediate Feedback and Repeated Pronunciation Practice

The first analytical area concerned how learners experienced the availability of immediate feedback during pronunciation practice. The researcher examined whether participants described using the assistant to repeat words, phrases, sentences, or short spoken responses and whether the feedback influenced their subsequent attempts. Particular attention was given to the relationship between receiving feedback and deciding what to do next, rather than assuming that repetition automatically represented improvement. Evidence for this theme should demonstrate whether learners used automated responses to identify difficulties, monitor their attempts, or sustain practice when teacher assistance was unavailable. Observation notes and digital artefacts may help establish whether participants’ descriptions corresponded with their actual interaction patterns. The final report should include a representative excerpt from a participant who described a specific feedback episode, followed by an interpretation of what that episode revealed about the learner’s approach to pronunciation practice. This analysis would clarify whether immediate feedback functioned primarily as a source of information, a prompt for repeated practice, or a combination of both.

4.3. Increased Awareness of Pronunciation Difficulties

The second analytical area focused on learners’ awareness of pronunciation difficulties and their understanding of the relationship between spoken production and automated recognition. The analysis examined whether participants identified particular sounds, words, stress patterns, or speech sequences that required further attention after using the application. It also considered whether learners distinguished between a pronunciation difficulty and a recognition problem caused by the limitations of the technology. Such distinctions are important because an automated transcript does not necessarily provide a complete or reliable diagnosis of every pronunciation feature. Where participants reported changes in their awareness, the analysis should identify the experiences that prompted those changes and explain how learners responded to the information. For example, a participant might have compared an intended sentence with the application’s transcription and then reconsidered a particular word. This example is illustrative and should only appear as a reported finding if supported by the actual dataset. The final analysis should also preserve accounts in which learners found the feedback difficult to interpret or did not perceive any meaningful change in their pronunciation awareness.

4.4. Trust, Recognition Inconsistencies, and Learner Responses

The third analytical area examined how learners evaluated the reliability of AI generated pronunciation feedback. Rather than treating trust as a fixed personal characteristic, the analysis considered whether participants’ confidence in the application varied according to the task, the feedback received, and their own understanding of the spoken expression. Relevant evidence included instances in which learners accepted a correction, repeated an utterance, questioned a recognition result, or sought an alternative explanation. The analysis also examined how learners distinguished between errors in their own speech and possible limitations of the recognition system. This distinction was central to understanding the relationship between automated assistance and learner judgment. Previous research by Zou et al. (2023) documented both the perceived potential of AI speech evaluation and the limitations learners encountered when using automated systems. These studies provide a relevant basis for interpreting the present findings, but they cannot substitute for evidence from the current participants. The final results should therefore explain how trust was established, challenged, or revised in the particular learning context under investigation.

Table 2. Suggested evidence matrix for analyzing responses to automated feedback.

Observed or reported response	Analytical question	Evidence required
Accepted the feedback	What made the suggestion credible to the learner?	Interview excerpt and relevant feedback record
Repeated the utterance	What motivated another attempt?	Observation note or practice record
Questioned the feedback	What created uncertainty?	Interview excerpt and digital artefact
Sought additional guidance	Why was automated feedback insufficient?	Interview account or observation note

Table 2. Suggested evidence matrix for analyzing responses to automated feedback.

Observed or reported response	Analytical question	Evidence required
Accepted the feedback	What made the suggestion credible to the learner?	Interview excerpt and relevant feedback record
Repeated the utterance	What motivated another attempt?	Observation note or practice record
Questioned the feedback	What created uncertainty?	Interview excerpt and digital artefact
Sought additional guidance	Why was automated feedback insufficient?	Interview account or observation note

4.5. Learner Agency and Independent Learning Strategies

The fourth analytical area concerned learner agency, particularly how participants made decisions about pronunciation practice rather than simply following the application’s suggestions. The analysis examined whether learners selected particular words for further practice, changed their approach after receiving feedback, consulted alternative resources, or decided when additional repetition was no longer useful. It also considered how learners managed uncertainty and whether they relied on the application alone or combined automated assistance with their existing language learning strategies. These processes offered a way to examine learner autonomy as a set of situated decisions rather than as a general perception of independence. Where the data supported such interpretations, the researcher could distinguish between strategic use of feedback and mechanical repetition without clear evaluation. Contrasting accounts would also be important because learners might differ in their confidence, familiarity with pronunciation learning, and ability to interpret system responses. The final thematic account should explain these differences and illustrate how participants exercised agency in practice, including situations in which they remained dependent on automated judgments or struggled to determine an appropriate next step.

4.6. Teacher Guidance and the Limits of Automated Assistance

The fifth analytical area examined learners’ perceptions of the relationship between AI powered pronunciation assistants and teacher guidance. The analysis considered whether participants found automated feedback sufficient for straightforward practice but required additional explanation for more complex pronunciation difficulties. Particular attention was given to learners’ accounts of stress, rhythm, intonation, connected speech, and the relationship between pronunciation and communicative meaning. It also examined whether participants viewed the application as a supplementary resource or as a substitute for interaction with a teacher. This distinction was relevant to the study’s pedagogical focus because the usefulness of automated feedback depended not only on its availability but also on learners’ ability to interpret and apply it appropriately. UNESCO (2023) emphasized the importance of maintaining human agency and pedagogical oversight in educational uses of AI. In the present study, that principle could be explored through concrete accounts of situations in which learners sought clarification, compared system feedback with teacher explanations, or encountered limitations that the application could not resolve. Any conclusions about the complementary role of teachers should be grounded in participants’ accounts and observed learning practices rather than presented as an assumption.

4.7. Cross Theme Analysis and Triangulation

After developing and reviewing the candidate themes, the researcher compared the interview accounts, observation notes, and digital artefacts to examine how the different sources supported, complicated, or contradicted one another. This stage was intended to produce an integrated interpretation rather than three separate summaries of the data collection instruments. For example, an interview account describing frequent independent practice could be compared with available practice records, while a participant’s concern about recognition accuracy could be examined alongside the relevant feedback screenshot. Agreement across sources would strengthen the contextual interpretation, whereas differences would provide opportunities to investigate how participants understood their own learning experiences. The analysis also considered negative or divergent cases, including participants who reported limited benefits, questioned the usefulness of the application, or preferred teacher feedback. Such cases were important because they prevented the findings from presenting AI supported pronunciation learning as a uniformly positive experience. Consistent with reflexive thematic analysis, the researcher treated the development of themes as an interpretive process and documented the reasoning behind analytical decisions (Braun & Clarke, 2021). The final account should identify the evidence supporting each theme and explain how the themes collectively addressed the research questions.

Table 3. Proposed cross source analysis matrix.

Research question	Primary data source	Supporting data source	Analytical outcome
How did learners experience AI pronunciation assistants?	Semi structured interviews	Observation notes	Experiences and perceived usefulness
How did learners interpret and respond to feedback?	Interviews and feedback records	Practice observations	Trust, uncertainty, and response strategies
What benefits, challenges, and strategies did learners identify?	Interviews	Digital artefacts and field notes	Enabling conditions, limitations, and learning strategies

Table 3. Proposed cross source analysis matrix.

Research question	Primary data source	Supporting data source	Analytical outcome
How did learners experience AI pronunciation assistants?	Semi structured interviews	Observation notes	Experiences and perceived usefulness
How did learners interpret and respond to feedback?	Interviews and feedback records	Practice observations	Trust, uncertainty, and response strategies
What benefits, challenges, and strategies did learners identify?	Interviews	Digital artefacts and field notes	Enabling conditions, limitations, and learning strategies

4.8. Summary of the Findings and Answers to the Research Questions

The final synthesis should bring together the themes that were demonstrably supported by the dataset and explain their relationship to the three research questions. The first question requires an account of how learners experienced AI powered pronunciation assistants during their practice, including the activities they performed and the meanings they attached to the experience. The second question requires a closer examination of how learners interpreted automated feedback, including their decisions to accept, question, or act on particular suggestions. The third question requires an account of the benefits, limitations, and strategies identified by participants, with attention to differences across their experiences. Rather than reporting that all participants benefited from the technology, the analysis should distinguish recurring patterns from individual experiences and explain any contradictions that emerged. The final section should include a small number of carefully selected participant quotations, each linked to a clearly defined analytical point. It should conclude with a concise synthesis of what the qualitative evidence revealed about the relationship between automated feedback, learner agency, and teacher guidance in the specific EFL context. This synthesis should remain descriptive and interpretive, leaving broader implications and comparisons with previous scholarship for the Discussion section.

5. Discussion

5.1. AI Powered Pronunciation Assistants for Sustained Pronunciation Practice

This study examines learners’ experiences with AI powered pronunciation assistants, focusing on feedback, pronunciation awareness, trust, learner agency, and teacher guidance. Although these tools enable repeated practice and immediate feedback, their value depends on learners’ ability to interpret responses and make informed decisions. Consistent with Zou et al. (2023) and Liu et al. (2025), the discussion emphasizes learning processes rather than assuming that repeated use automatically improves pronunciation. This perspective highlights the importance of understanding how learners incorporate automated assistance into their individual learning routines. It also distinguishes access to technology from meaningful engagement with pronunciation practice.

5.2. Automated Feedback, Pronunciation Awareness, and the Interpretation of Errors

Automated feedback can raise pronunciation awareness, but recognition errors may reflect either production difficulties or system limitations. Learners must evaluate whether feedback matches their intended speech and determine appropriate responses. Repeated practice alone does not demonstrate improved pronunciation. Teacher guidance remains important in distinguishing useful feedback from misleading results and encouraging critical reflection. Learners may benefit from comparing automated responses with their own listening judgments and other learning resources. Such activities can help them develop a more reflective approach to identifying and addressing pronunciation difficulties (Santiana et al., 2021; Syafryadin et al., 2024).

5.3. Learner Agency, Trust, and Independent Learning

Learner agency involves deciding whether to accept automated corrections, repeat practice, or seek additional support. This perspective aligns with UNESCO (2023), which emphasizes human judgment in educational AI, and Liu et al. (2025), who highlight the need for stronger theoretical grounding in ASR research. Teachers can promote informed decision making by encouraging learners to question, compare, and evaluate automated feedback rather than accept it uncritically. Learners’ trust in these tools may depend on their experiences with recognition accuracy and feedback consistency. Developing critical judgment is therefore essential for independent learning without excessive reliance on automated evaluations (Marzuki, 2016; Marzuki, 2017; Santiana & Marzuki, 2022)

5.4. Teacher Mediation and the Pedagogical Integration of AI

Pronunciation involves stress, rhythm, intonation, and connected speech, which cannot be fully assessed through recognition scores. AI assistants can extend practice opportunities, while teachers provide contextual explanations and address complex learning needs. In Indonesian EFL settings, this complementary approach may support practice beyond classroom hours, provided learners have adequate access and guidance. Conclusions about its value should be grounded in participants’ actual experiences. Teachers can also help learners identify recurring difficulties and select appropriate strategies for further practice. Integrating AI into structured learning activities may make pronunciation practice more purposeful and communicatively relevant (Muthmainnah et al., 2022; Muthmainnah, Marzuki et al., 2024).

5.5. Theoretical and Practical Implications, Limitations, and Future Research

This study connects automated feedback, learner agency, and teacher mediation in AI assisted pronunciation learning. Its practical implications include integrating individual practice, critical reflection, and teacher guidance. However, a qualitative case study involving 18 students from one institution offers contextual insights rather than statistical generalizability. Interviews, observations, and digital artefacts cannot independently establish measurable pronunciation gains. Future research should compare tools and learning contexts, examine experiences over time, and incorporate acoustic analysis or intelligibility assessments. Further studies could also investigate how proficiency levels and digital familiarity influence learners’ responses to automated feedback. These comparisons would provide a clearer understanding of the conditions under which pronunciation assistants can support meaningful learning (Marzuki, 2019c; Marzuki, & Kuliahana, 2021; Marzuki & Santiana, 2022).

5.6. Overall Contribution to AI Supported Pronunciation Learning

AI powered pronunciation assistants are best understood as supplementary learning resources whose value depends on feedback interpretation, technological reliability, and pedagogical support. Building on Zou et al. (2023) and Liu et al. (2025), this study highlights the importance of examining learners’ responses and decisions during practice. Effective integration requires a balance between technological assistance, learner autonomy, teacher guidance, and communicative needs. By foregrounding learners’ experiences, this perspective shifts attention from technological capabilities alone to how automated feedback is understood and used in real learning contexts. This contribution may inform more critical and pedagogically grounded approaches to AI supported pronunciation instruction.

6. Conclusion

This study highlights the potential of AI powered pronunciation assistants to support EFL learning by providing immediate feedback, encouraging repeated practice, and promoting learner autonomy. However, the usefulness of these tools depends on learners’ ability to interpret automated feedback critically, as speech recognition results may not always accurately reflect pronunciation quality. The findings framework therefore emphasises the complementary roles of AI technology and teacher guidance in helping learners identify pronunciation difficulties, develop independent learning strategies, and improve their spoken communication. Rather than replacing teachers, AI powered pronunciation assistants are best integrated as supplementary resources within pedagogically guided learning activities. Future research should examine learners’ experiences over longer periods and combine qualitative insights with objective pronunciation assessments to determine how AI supported practice contributes to pronunciation development and communicative competence across different EFL contexts.