Integrating artificial intelligence in EFL speaking instruction: Opportunities and challenges at the Industrial University of Ho Chi Minh City

ABSTRACT:

This study examines the integration of artificial intelligence (AI) into EFL speaking instruction at the Industrial University of Ho Chi Minh City (IUH). After ten weeks of AI-supported learning, students showed significant improvements in pronunciation, fluency, confidence, and learner autonomy, while teachers reported higher classroom engagement and more effective formative assessment. Students also experienced reduced speaking anxiety and increased willingness to communicate. However, challenges related to technology access, accent bias, overreliance on AI, and the need for pedagogical mediation were identified. The findings suggest that AI can effectively enhance EFL speaking skills when carefully integrated into communicative teaching frameworks under educator guidance.

Keywords: artificial intelligence, EFL speaking instruction, learner autonomy, pronunciation feedback, Vietnamese higher education.

1. Introduction

In recent years, Artificial Intelligence (AI) has moved from being a futuristic concept to an integral component of educational innovation worldwide. The rapid development of AI-driven technologies has begun to transform how teachers deliver lessons and how students engage with learning materials. In language education, particularly in English as a Foreign Language (EFL) settings, AI has shown great potential in supporting students’ language acquisition through intelligent tutoring systems, automatic feedback, and personalized learning environments (Luckin et al., 2016; Zawacki-Richter et al., 2019).

Among the four language skills - listening, speaking, reading, and writing -speaking remains one of the most challenging for learners in non-English-speaking contexts. Speaking requires real-time processing of linguistic, cognitive, and social information. Unlike reading or writing, oral communication demands immediate response, appropriate pronunciation, and interactional strategies. As Brown (2001) emphasizes, speaking competence reflects not only linguistic knowledge but also communicative and pragmatic competence, which are difficult to develop in classroom-based learning environments.

In Vietnam, although English has become a compulsory subject from primary to tertiary levels, the majority of university students, especially non-English majors, still struggle to communicate effectively in spoken English. Students at Industrial University of Ho Chi Minh City (IUH) are no exception. They often possess sufficient grammatical knowledge and reading comprehension but display low oral fluency, limited vocabulary for spontaneous communication, and poor pronunciation accuracy. Many learners report feelings of anxiety and embarrassment when speaking in front of peers, which leads to avoidance behaviors and limited progress (Nguyen & Tran, 2022).

The traditional English-speaking classes at IUH are typically teacher-centered, focusing on repetition, dialogue memorization, and scripted role-plays rather than authentic interaction. Furthermore, class sizes of 40-50 students restrict teachers’ ability to provide individual feedback. Consequently, students rarely receive the timely corrective feedback essential for developing speaking skills.

In this context, AI presents a new pedagogical opportunity. AI-powered applications such as ELSA Speak, ChatGPT, Google Assistant, and Duolingo Max can analyze students’ pronunciation, suggest improvements, and simulate real-life communication scenarios. These technologies have the potential to personalize speaking practice, allowing students to practice independently and receive immediate, individualized feedback-something difficult to achieve in traditional classrooms.

However, despite these advantages, integrating AI into speaking instruction in Vietnam is still at an experimental stage. Many teachers are uncertain about how to integrate AI tools effectively into their teaching methodology or align them with communicative language teaching (CLT) principles. Others express concerns about AI’s accuracy, the authenticity of AI-generated dialogues, and the risk of overreliance on technology that may reduce genuine human communication.

Therefore, this research aims to examine how AI can be strategically and pedagogically integrated into EFL speaking classes at IUH University to enhance learners’ oral proficiency and motivation while maintaining the human-centered nature of teaching. Specifically, it investigates (1) how AI tools influence students’ speaking performance, (2) how students and teachers perceive the use of AI in EFL speaking instruction, and (3) what challenges and ethical implications arise during AI integration.

This study is significant for several reasons. Firstly, it contributes to the limited body of research on AI-assisted speaking instruction in Vietnamese higher education, especially among non-English-major students. Secondly, it provides empirical evidence and classroom-based practices that may inform future curriculum development and teacher training programs. Finally, it highlights the need to balance technological innovation with pedagogical integrity, ensuring that AI serves as an assistant-not a replacement-for human teachers.

2. Problem statement

Although the integration of English into university curricula in Vietnam has been extensive, the oral proficiency level of many students remains below expectations. At IUH University, internal assessments and informal teacher observations reveal that most students perform well in written and reading exams but perform poorly in speaking tests. Typical problems include:

- Limited fluency and confidence: Students hesitate, pause frequently, and rely heavily on written preparation before speaking.

- Pronunciation and intonation errors: Many learners struggle to distinguish between similar sounds or stress patterns, leading to misunderstandings.

- Low motivation: Speaking activities are often perceived as stressful rather than enjoyable.

These issues are partly caused by instructional limitations. Due to large class sizes, instructors can only allocate limited time to oral practice. As a result, speaking lessons often focus on mechanical drills rather than communicative tasks. Feedback on pronunciation and fluency is usually delayed or superficial, preventing learners from noticing and correcting their mistakes.

In this context, AI technology could bridge the gap between classroom teaching and individualized practice. AI-based tools can provide instant pronunciation evaluation, visual feedback, and interactive dialogue simulations. For example, ELSA Speak rates users’ speech accuracy and offers corrective feedback on stress and rhythm, while ChatGPT allows learners to engage in unlimited English conversations in various topics. These tools may help students practice more frequently and autonomously, leading to improved fluency and pronunciation accuracy.

However, several challenges hinder the effective integration of AI in teaching speaking at IUH University:

1) Limited awareness and training among teachers: Many instructors have not received professional development on how to use AI pedagogically. They may view AI as an optional supplement rather than an instructional tool that can be embedded in lesson planning.

2) Technological constraints: Not all students have access to reliable internet or compatible devices, particularly in large classes. This reduces the inclusivity of AI-based activities.

3) Pedagogical misalignment: Some AI applications focus on isolated pronunciation drills rather than communicative use of language, which may conflict with CLT approaches emphasizing interaction and meaning-making.

4) Ethical and accuracy concerns: AI systems are not always sensitive to non-native accents. Vietnamese-accented English may be misjudged as incorrect, leading to frustration among learners. Additionally, overreliance on AI-generated content may limit students’ creativity and critical thinking.

5) Lack of institutional policy: IUH currently lacks a clear policy or guideline on the pedagogical use of AI in classrooms, leaving decisions to individual teachers’ discretion.

Given these realities, there is a pressing need to explore how AI can be integrated effectively and responsibly into EFL speaking classes at IUH. The purpose of this study is not merely to introduce AI as a technological innovation but to evaluate its pedagogical impact-specifically, how it enhances learners’ speaking performance, engagement, and autonomy within a Vietnamese university context.

3. Research at the IUH University

3.1. Institutional background

The Industrial University of Ho Chi Minh City (IUH) is one of the largest multidisciplinary universities in southern Vietnam, with more than 60,000 students across various faculties such as engineering, economics, hospitality, and information technology. English is a mandatory subject for all non-English-major students, taught over three semesters in the general English program.

The curriculum focuses on four language skills - listening, speaking, reading, and writing-but due to time constraints and class sizes (often 40–50 students), speaking receives the least attention. Assessments traditionally emphasize written grammar and reading comprehension rather than oral communication, which reduces students’ motivation to develop speaking fluency.

3.2. Students’ English proficiency and learning habits

Most IUH students enter university at the A2–B1 level on the CEFR scale. While they can understand textbook dialogues, they struggle to respond spontaneously or maintain a conversation. Interviews conducted before the study revealed that:

- 83% of students felt anxious speaking English in class.

- 71% preferred practicing alone rather than in groups.

- 68% believed their pronunciation was “poor” or “unintelligible.”

Students expressed a strong desire for personalized feedback and flexible practice time, something not feasible under traditional teaching conditions.

3.3. Artificial intelligence - assisted speaking initiative

In the 2024-2025 academic year, IUH’s English Department launched a pilot project integrating AI tools into the speaking curriculum. The project used a blended approach combining:

- ELSA Speak for pronunciation and intonation practice.

- ChatGPT for interactive conversation and role-plays.

- Google Speech Recognition for oral assessments.

Teachers were trained to design AI-supported speaking tasks, allowing students to practice at home and apply their learning during classroom communicative activities.

4. Research design and methodology

Participants included: 120 non-English-major students (aged 18-21) from three faculties: Business Administration, Electrical Engineering, and Tourism. 5 English lecturers with 5-12 years of teaching experience.

Research instruments: Pre-and post-speaking tests rated by three trained assessors using the CEFR criteria (fluency, pronunciation, vocabulary, and coherence). Questionnaires on student attitudes toward AI-based speaking practice. Semi-structured interviews with teachers. Usage data from ELSA Speak and ChatGPT activity logs.

Research Procedure:

+ Week 1-2: Baseline test, orientation on AI tools.

+ Week 3-12: Weekly AI-based speaking assignments integrated into regular lessons.

+ Week 13-15: Post-test, reflection, and interviews.

Data Analysis: Quantitative data were analyzed using descriptive statistics and paired-sample t-tests. Qualitative data were coded into themes: motivation, autonomy, feedback quality, and challenges.

5. Reseults

The integration of AI tools into EFL speaking classes at IUH University generated a broad range of positive learning outcomes, both measurable and perceptual. This section details the observed improvements in speaking performance, motivation, confidence, and learner autonomy, alongside notable challenges identified throughout the project.

5.1. Improvement in speaking proficiency

The quantitative data revealed remarkable progress in all key speaking criteria. Students’ pronunciation, fluency, and lexical range improved significantly, as illustrated in pre- and post-test results. Beyond numerical growth, the quality of speech delivery changed perceptibly: students spoke with smoother pacing, more natural rhythm, and clearer articulation.

Teachers noted that, by Week 8, students were able to maintain 2-3 minute conversations without relying on pre-written notes. Before AI integration, most students struggled to sustain speech for even 45 seconds. This suggests that AI-enhanced repetitive practice built both linguistic competence and speech stamina.

Example: A business administration student reported, ‘Before, I always forgot what to say after two sentences. After using ChatGPT, I practiced how to keep the conversation going. Now I can respond naturally and even ask questions back.’

5.2. Pronunciation accuracy and intonation

Through ELSA Speak’s real-time feedback, students became more aware of phonetic details such as stress, rhythm, and connected speech. The software displayed color-coded sound accuracy and allowed learners to compare their voice waveforms with native models. This visualization of sound was particularly effective for Vietnamese learners, who often have difficulty perceiving stress and intonation patterns. After consistent practice, the average pronunciation score (on a 5-point rubric) rose from 2.5 to 3.4.

Some students described how AI feedback clarified specific errors teachers had long struggled to address in large classes. For example, mispronunciation of final consonants (“work” → “wor”) dropped noticeably. One learner commented: ‘ELSA made me realize I often miss ending sounds. Now I pay attention every time I speak.’

5.3. Growth in confidence and speaking willingness

Survey results showed that 78% of students felt “much more confident” after AI-assisted speaking activities. They appreciated being able to practice without embarrassment or fear of being corrected in front of peers.

Students repeatedly mentioned that AI provided a safe, judgment-free environment for experimenting with language. ‘I can repeat many times with ELSA or ChatGPT until I sound right. A teacher might get tired, but AI never does,’ one student explained.

As confidence improved, classroom participation also rose. Teachers observed that formerly silent students began volunteering for role-plays and oral presentations.

5.4. Learner autonomy and study habits

The data logs showed that 64% of students practiced with AI tools outside of assigned class hours, often late at night or during weekends. This demonstrates the shift from teacher-dependent learning to self-directed learning.

ELSA’s gamified progress charts encouraged goal-setting, while ChatGPT’s conversational flexibility allowed endless practice in students’ chosen topics - such as travel, technology, or career interviews.

This autonomous learning culture is a notable breakthrough in Vietnamese higher education, where learners are typically accustomed to teacher-led instruction.

5.5. Teachers’ observations

Teachers unanimously agreed that AI tools enhanced lesson efficiency and provided richer data for formative assessment. For instance, ELSA reports allowed teachers to identify class-wide weaknesses (like “// sound confusion”) and tailor corrective activities accordingly. However, they also noted several concerns:

- Some students overrelied on AI phrasing, leading to memorized speech patterns.

- AI occasionally misjudged Vietnamese-accented but intelligible English.

- Not all students had equal access to reliable internet connections.

Despite these limitations, all instructors agreed that AI integration is pedagogically beneficial when used strategicallywithin communicative frameworks.

5.6. Emerging learning behaviors and communication strategies

An important finding not fully captured by quantitative data is the evolution of students’ communication strategies. Throughout the ten-week AI integration, students began adopting techniques that reflected deeper metacognitive control over their speaking process. For example, several students reported that after using ChatGPT, they learned to paraphrase ideas, use fillers (“let me think,” “you know”), and employ interactive repair strategies such as rephrasing when not understood.

This suggests that exposure to AI-generated conversation models helped students internalize discourse management techniques used by proficient speakers. Unlike textbook dialogues, AI conversations offered authentic examples of turn-taking, hedging, and politeness strategies.

Additionally, teachers observed that students gradually shifted from form-focused learning (correcting single sounds or grammar) to meaning-focused communication (maintaining message flow despite errors). This is a significant pedagogical transformation consistent with the Communicative Approach and Krashen’s Input Hypothesis (1982), which emphasizes comprehensible input and low anxiety as conditions for language acquisition.

The project also uncovered an improvement in self-monitoring ability. Students started evaluating their own pronunciation without explicit teacher instruction. Some even created “AI journals,” recording their weekly progress based on ELSA feedback graphs. This self-observation indicates a high degree of metalinguistic awareness, a key factor in long-term language retention (Ellis, 2008).

6. Discussion

This section critically interprets the findings, highlighting the pedagogical and psychological mechanisms through which AI facilitated speaking improvement and reshaped classroom dynamics at IUH.

6.1. Pronunciation development through AI-enhanced feedback

Pronunciation improvement was the most tangible outcome. Traditional instruction relies on the teacher’s auditory judgment, which can be inconsistent or limited by time. AI tools, in contrast, offer consistent, individualized, and multimodal feedback. Students could see their pronunciation errors, not merely hear them, which aligns with dual-coding theory (Paivio, 1986) - suggesting that visual-auditory integration enhances retention.

At IUH, students reported that repeating AI feedback multiple times improved muscle memory for articulation. In essence, AI served as a “mirror tutor,” providing precise and non-judgmental correction, reinforcing students’ metacognitive awareness of pronunciation.

However, AI feedback should be contextualized by teachers. Overly mechanical mimicry may lead to robotic speech. Thus, educators should guide students to use AI feedback selectively-focusing on intelligibility rather than perfect accent imitation.

6.2. AI and the development of fluency

Fluency involves speed, smoothness, and coherence of speech. ChatGPT’s conversational features enabled IUH students to practice extended turns in interaction, which traditional textbooks rarely provide.

AI dialogues simulate authentic communicative pressure - requiring immediate responses – thus improving students’ real-time processing skills. Repeated conversational practice helped learners internalize formulaic expressions (“That’s a good point, but I think…”) and discourse markers (“actually,” “in my opinion,” “as a result”), which contribute to natural flow.

Moreover, ChatGPT’s ability to generate dynamic replies encouraged adaptive thinking. Unlike scripted tasks, students could not predict the AI’s response, forcing them to listen actively and improvise - an essential aspect of communicative competence.

6.3. Cognitive and affective benefits

AI created an emotionally safe space for learners who typically experience communication anxiety. Students could practice privately without fear of negative evaluation (Horwitz et al., 1986). This reduced affective barriers and increased willingness to communicate (WTC).

The repetitive yet supportive nature of AI practice triggered positive reinforcement cycles: success in AI feedback boosted confidence → increased practice → improved performance → higher self-efficacy.

Cognitively, AI allowed individualized pacing. Faster learners could advance to complex dialogues, while slower learners repeated basic patterns until mastery. This level of personalization is impossible in large IUH classes.

6.4. Pedagogical transformation: Teacher’s role redefined

The teacher’s role evolved from “speaker” to “facilitator of communicative learning.” Instead of spending class time drilling pronunciation, teachers could analyze AI-generated data to focus on interactional aspects - such as turn-taking, tone, and pragmatics.

Teachers described themselves as coaches and interpreters of AI feedback, helping students contextualize automated scores and linking them to real-life communication goals.

This supports a human-AI partnership model where technology handles micro-level feedback, and the teacher fosters macro-level communication competence.

However, for sustainable impact, teachers require ongoing AI literacy training. They must understand not only how to operate tools but also how to align them with task-based language teaching (TBLT) and communicative language teaching (CLT) principles.

6.5. Socio-technical barriers

The study revealed several socio-technical challenges:

- Digital inequality: Some IUH students lacked personal smartphones or stable connections, limiting participation.

- Accent and bias: AI trained on Western accents often rated local speech unfairly low, affecting student morale.

- Overreliance risk: Some learners began copying ChatGPT-generated answers without genuine comprehension.

These findings emphasize that technological access and digital ethics must accompany pedagogical innovation. Universities need clear policies on privacy, fairness, and responsible AI use.

6.6. Cultural shift toward autonomous learning

Perhaps the most transformative change observed was cultural: AI shifted students’ learning mindset from passive reception to active construction.

Vietnamese classrooms are often hierarchical; students wait for teacher approval. AI’s constant availability and patience empowered learners to take initiative.

As a result, students began viewing speaking as a skill to train daily, not a test to pass occasionally. This autonomy-oriented culture, once cultivated, can outlast the technology itself-representing a sustainable pedagogical evolution.

6.7. AI, social learning, and peer collaboration

Although AI tools mainly supported individual practice, they unexpectedly enhanced peer interaction. Students began comparing AI scores and sharing strategies to improve pronunciation accuracy or conversational coherence. In group work, they used ChatGPT collectively to generate discussion prompts or role-play situations, then adapted them into personalized dialogues.

This hybrid collaboration fostered a community of practice (Lave & Wenger, 1991) where technology became a social mediator rather than an isolating tool. Teachers noted that AI acted as a conversation starter, helping shy students enter discussions with prepared phrases.

One teacher remarked: “Even the quietest students came to class with ideas they practiced with ChatGPT. It gave them something to say, which is half the battle in teaching speaking.”

Moreover, the AI-generated materials were often linguistically richer and contextually varied, exposing students to new vocabulary and expressions. As a result, the classroom environment became more dialogic and inclusive, bridging gaps between high- and low-proficiency learners.

6.8. Cognitive overload and balancing human interaction

Despite the positive outcomes, AI tools also risked cognitive overload. Beginners sometimes found ChatGPT’s responses too fast or complex. Without teacher scaffolding, they became discouraged. Teachers thus learned to curate AI content carefully - adjusting prompt difficulty and encouraging students to simplify AI outputs for peer understanding.

This illustrates a crucial pedagogical principle: AI must be mediated by human empathy. Teachers serve as cognitive filters, adjusting technological input to match learners’ readiness. Otherwise, excessive exposure to AI’s linguistic complexity can overwhelm learners’ working memory and hinder speaking confidence.

In other words, AI brings efficiency, but teachers preserve human sensitivity. The synergy between both yields optimal outcomes.

6.9. Institutional and policy implications

Beyond classroom-level findings, this study reveals the need for institutional readiness at IUH and similar universities. While students embraced AI enthusiastically, teachers emphasized the importance of:

- Technical infrastructure (strong Wi-Fi, classroom screens, AI-compatible devices).

- Professional development programs for digital pedagogy.

- Ethical guidelines concerning AI data storage, voice recording, and academic honesty.

Without such frameworks, the sustainability of AI integration is at risk. Institutions must view AI not as a short-term project but as part of strategic educational transformation - aligning with Vietnam’s national policy on digital education (MOET, 2024).

Furthermore, university administrators should encourage action research projects like this one, allowing teachers to explore, adapt, and publish empirical results. The AI initiative at IUH can serve as a model of bottom-up innovation, demonstrating how local experimentation drives national reform.

8. Conclusions

The integration of AI into speaking instruction for EFL learners at IUH University represents a paradigm shift in both pedagogy and learner psychology. The findings go beyond simple improvement in pronunciation or fluency – they reveal a transformation of attitudes, autonomy, and classroom culture.

AI technologies, when pedagogically guided, provide a scalable solution to chronic challenges in Vietnamese EFL classrooms: large class sizes, limited teacher feedback time, and students’ communication anxiety. Through automated speech feedback and conversation simulation, learners receive individualized attention once thought impossible in resource-limited settings.

However, the success of AI integration ultimately depends on human mediation. Teachers remain irreplaceable as motivators, ethical guides, and cultural interpreters. Their empathy, contextual understanding, and ability to foster interpersonal connection ensure that AI remains a tool for empowering human communication, not replacing it.

This human-AI balance is essential for sustainability. Without teacher agency, technology risks turning language learning into mechanical score-chasing; without AI, traditional methods may remain inefficient and demotivating. Together, they form a hybrid pedagogical ecosystem where technology amplifies teacher effectiveness and extends learning beyond classroom walls.

Looking ahead, AI integration should evolve toward localization and personalization. Future systems should adapt to Vietnamese-English phonology, offer feedback in learners’ native language when necessary, and respect cultural nuances in communication styles. Further longitudinal studies across multiple institutions can track whether the improvements observed at IUH persist and generalize across other EFL contexts.

Finally, this research contributes to the growing understanding that AI is not just a tool for efficiency but a catalyst for educational equity. By providing individualized oral practice to every student, regardless of background or proficiency level, AI helps democratize access to high-quality language learning - a goal at the heart of IUH’s educational mission and Vietnam’s broader vision for digital transformation in education.

REFERENCES:

Brown, H. D. (2001). Teaching by principles: An interactive approach to language pedagogy (2nd ed.). Pearson Education.

Deci, E. L., & Ryan, R. M. (2000). The “What” and “Why” of goal pursuits: Human needs and the self-determination of behavior. Psychological Inquiry, 11(4), 227-268.

Ellis, R. (2008). The study of second language acquisition (2nd ed.). Oxford University Press.

Holmes, W., Bialik, M., & Fadel, C. (2022). Artificial intelligence in education: Promises and implications for teaching and learning. Center for Curriculum Redesign.

Horwitz, E. K., Horwitz, M. B., & Cope, J. (1986). Foreign language classroom anxiety. The Modern Language Journal, 70(2), 125-132.

Krashen, S. D. (1982). Principles and practice in second language acquisition. Pergamon Press.

Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. Cambridge University Press.

Luckin, R., Holmes, W., Griffiths, M., & Forcier, L. B. (2016). Intelligence unleashed: An argument for AI in education.Pearson.

Ministry of Education and Training of Vietnam. (2024). National digital transformation strategy for higher education 2024–2030. Hanoi: MOET Publishing.

Nguyen, T. H., & Tran, M. Q. (2022). Factors affecting Vietnamese EFL students’ speaking anxiety. Vietnam Journal of Education Research, 15(2), 45–58.

Paivio, A. (1986). Mental representations: A dual coding approach. Oxford University Press.

Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Harvard University Press.

Wang, Y., & Lin, J. (2022). AI-supported speaking instruction in college EFL classrooms: A quasi-experimental study. Asian EFL Journal, 24(3), 112-133.

Zawacki-Richter, O., Marín, V. I., Bond, M., & Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education. International Journal of Educational Technology in Higher Education, 16(1), 39-57.

ỨNG DỤNG TRÍ TUỆ NHÂN TẠO TRONG GIẢNG DẠY

KỸ NĂNG NÓI TIẾNG ANH NHƯ MỘT NGOẠI NGỮ (EFL):

NGHIÊN CỨU TẠI TRƯỜNG ĐẠI HỌC CÔNG NGHIỆP THÀNH PHỐ HỒ CHÍ MINH

 

• VÕ THỊ MINH NGÂN

Trường Đại học Công nghiệp Thành phố Hồ Chí Minh

 

TÓM TẮT:

Nghiên cứu này đánh giá việc tích hợp trí tuệ nhân tạo (AI) vào giảng dạy kỹ năng nói tiếng Anh (EFL) tại Trường Đại học Công nghiệp Thành phố Hồ Chí Minh (IUH). Sau mười tuần triển khai việc tích hợp AI vào hoạt động giảng dạy, các sinh viên ghi nhận sự cải thiện rõ rệt về độ chính xác phát âm, mức độ trôi chảy, sự tự tin và năng lực tự chủ trong học tiếng Anh; đồng thời, giảng viên cho thấy mức độ sinh viên tham gia lớp học cao hơn và việc đánh giá quá trình học tập của sinh viên trở nên hiệu quả hơn. Sinh viên cũng giảm bớt lo lắng khi nói và tăng cường mức độ sẵn sàng giao tiếp bằng tiếng Anh. Tuy nhiên, nghiên cứu cũng chỉ ra những thách thức liên quan đến khả năng tiếp cận công nghệ, thiên lệch giọng nói, nguy cơ phụ thuộc quá mức vào AI và yêu cầu về sự điều phối sư phạm. Kết quả cho thấy AI có thể nâng cao hiệu quả giảng dạy kỹ năng nói tiếng Anh khi được tích hợp một cách thận trọng trong các khung giảng dạy giao tiếp dưới sự hướng dẫn của giảng viên.

Từ khoá: trí tuệ nhân tạo, giảng dạy kỹ năng nói tiếng Anh như một ngoại ngữ, tính tự chủ của người học, phản hồi về phát âm, giáo dục đại học tại Việt Nam.

[Tạp chí Công Thương - Các kết quả nghiên cứu khoa học và ứng dụng công nghệ số 1 tháng 1 năm 2026]