Artificial Intelligence Implementation in Teaching English as a Foreign Language: A Qualitative Research Synthesis

Introduction

English proficiency is now widely recognized as essential in academic and professional domains (Mohamed, 2024), and educators continually seek to refine English language teaching methodologies, with technological advancements driving a shift from traditional to contemporary approaches in language education (Alrajhi, 2024; Ebadi & Amini, 2022). Furthermore, recent innovations in artificial intelligence (AI) have significantly enhanced the accessibility and effectiveness of English as a foreign language (EFL) teaching and learning resources (Hwang et al., 2012). Despite potential challenges associated with integrating such technologies, their implementation can foster more significant learner interaction and independence (Chen et al., 2022). Specifically, contemporary AI-powered learning tools are designed to cater to individual learning styles, creating interactive learning environments enriched with multimedia resources (Mohamed, 2024). As a result, computer-assisted language learning (CALL) has evolved in the modern era by integrating AI, offering innovations such as speech recognition technology to improve speaking skills (Zou et al., 2023). Moreover, AI can revolutionize language learning by automating tasks like grading and providing personalized feedback (Tai, 2022). These AI-driven platforms also promote inclusivity and enhance learner motivation by adapting to individual needs and simulating real-world scenarios (Fakher Ajabshir, 2023). Therefore, these AI-powered tools ultimately contribute to increased motivation, engagement, and language acquisition (Fakher Ajabshir, 2023; Han & Lee, 2022). However, these transformative changes necessitate a combination of creative vision and expert guidance to ensure their effective implementation (Zhang et al., 2024).

The effective integration of AI into EFL curricula depends on a thorough understanding of established frameworks that explain user acceptance of technology. Central to this understanding is the Technology Acceptance Model (TAM) (Figure 1), a foundational concept introduced by Davis (1989) that remains pivotal in information systems research. TAM offers a robust, theory-driven framework for analyzing drivers of technology adoption across diverse contexts and populations (Davis, 1989; Wei et al., 2025).

Rooted in the Theory of Reasoned Action (TRA) and the Theory of Planned Behavior (TPB), TAM identifies two key constructs that shape user acceptance. The first is perceived usefulness (PU), defined as the belief that technology enhances performance. The second is perceived ease of use (PEOU), which refers to the perception that technology is effortless to operate (Davis, 1989). These constructs directly influence users’ attitudes toward technology, subsequently shaping their behavioral intention (BI) and leading to actual system use (ASU) (Chen et al., 2024; Mata et al., 2024). For educators and administrators aiming to implement AI in EFL settings, TAM provides critical insights into the psychological and practical factors driving technology adoption.

The strength of TAM lies in the dynamic relationship between PU and PEOU. Empirical studies demonstrate that PEOU not only directly informs user attitudes but also indirectly enhances PU, with PU mediating the link between PEOU and attitude (Rejali et al., 2024). This interconnectedness strengthens TAM's predictive validity, as evidenced by its application in online learning, banking systems, and AI tools (Martinez et al., 2024; Wistedt, 2024).

Further refining the model, Venkatesh and Davis (2000) positioned behavioral intention (BI) as a critical intermediary between attitude and ASU. This adjustment clarifies the pathway from perception to action, underscoring TAM’s adaptability. For instance, Huang and Mizumoto (2024) confirmed that PEOU reinforces both PU and attitude, while PU independently strengthens attitude, collectively driving ASU. Such insights are invaluable for EFL contexts, where educators’ and learners’ perceptions of AI’s utility and usability are central to successful integration.

Figure 1. The Theoretical Framework of TAM (Davis et al., 1989; Mata et al., 2024)

Note: This figure is licensed under CC BY-NC 4.0 (https://creativecommons.org/licenses/by-nc-nd/4.0/).

While the Technology Acceptance Model (TAM) has garnered significant recognition, it has also faced criticism for its extrinsic focus and insufficient consideration of broader contextual variables. Kong et al. (2024) contend that TAM neglects macro-level influences, such as human, social, and organizational factors, significantly impacting technology adoption. This narrow emphasis on extrinsic motivators, particularly perceived usefulness (PU), has prompted scholars to advocate for integrating intrinsic factors like self-efficacy and subjective norms (Kong et al., 2024; Venkatesh, 1999). García et al. (2024) further stress the importance of social influence—including peer, familial, and educator input—in shaping behavioral outcomes, a dimension inadequately addressed in TAM's original framework. Additionally, TAM's applicability diminishes outside workplace settings, as its core constructs may fail to capture the complexities of diverse environments, such as educational contexts (Wei et al., 2025). These limitations highlight the need to critically assess TAM's boundaries when applying it to English as a Foreign Language (EFL) education, where pedagogical and social dynamics significantly influence AI implementation.

To address these critiques, later iterations of TAM have broadened its scope. For example, TAM2 (Venkatesh & Davis, 2000) and TAM3 (Venkatesh & Bala, 2008) incorporate constructs like subjective norms and facilitating conditions, aligning more closely with the holistic approach of the Theory of Planned Behavior (TPB) (Kleine et al., 2024). TAM3, in particular, advances the framework by dissecting mediating pathways through individual differences, social influences, and resource availability. It distinguishes between the antecedents of PU and perceived ease of use (PEOU): PU, akin to extrinsic motivation, is shaped by perceived benefits and external cues, while PEOU evolves from general computer beliefs refined through direct user experience (Venkatesh & Bala, 2008). This distinction underscores the role of hands-on interaction in solidifying PEOU, as users assess a system’s intuitiveness through practical engagement. For EFL educators introducing AI tools, this insight is critical for fostering user confidence.

TAM’s limitations have also spurred the development of integrative frameworks like the Unified Theory of Acceptance and Use of Technology (UTAUT) (Venkatesh et al., 2003). By synthesizing TAM with TPB and other theories, UTAUT incorporates performance expectancy, effort expectancy, social influence, and facilitating conditions, offering a comprehensive lens for analyzing technology acceptance (Kleine et al., 2024). Similarly, TPB enhances TAM by integrating subjective norms and perceived behavioral control, positing that behavioral intention (BI) arises from the interplay of attitude, social pressures, and perceived facilitators or constraints (Ajzen, 1985). However, TAM’s shortcomings remain pronounced in educational contexts, as it provides limited guidance on pedagogical integration or content delivery—key considerations in EFL curricula (García et al., 2024).

Despite these challenges, TAM remains vital for integrating AI into EFL education. It identifies critical factors influencing AI tool adoption, such as perceived usefulness, ease of use, and social influences. By leveraging TAM, stakeholders can anticipate challenges and align AI implementation with educational goals, ensuring technology enhances learning experiences and upholds the quality of EFL instruction.

Considering the historical context, the inception of AI in 1956, with its initial aim to replicate human intelligence (McCarthy, 2007), marked the beginning of a transformative influence on education. This process began to unfold from the 1960s onward. While technological applications in English language learning have a more extended history, exemplified by Joseph Weizenbaum's pioneering ELIZA program in 1966 (Chun, 2020), this early development already hinted at AI's revolutionary potential within language acquisition. Despite its inherent limitations, ELIZA provided a glimpse into future possibilities. Subsequently, AI-powered tools have significantly reshaped language education by offering learners personalized feedback and diverse learning resources (Russell & Norvig, 2016). Furthermore, projections indicate the development of even more sophisticated educational tools in the future (Lee, 2023; Mohamed & Shabaan, 2021; Wang et al., 2023). This advancement is evident in computer and mobile-assisted language learning through AI technologies (Zou et al., 2023). In addition, AI algorithms within educational systems facilitate personalized interventions tailored to individual learners (Demir & Güraksın, 2022; Jeon, 2022). Moreover, AI empowers teachers to make data-driven decisions regarding curriculum development and assessment through the analysis of student data (Teng et al., 2023). Concurrently, AI-based learning technologies offer individualized support and pinpoint students' strengths and weaknesses (Christudas et al., 2018; Goksel & Bozkurt, 2019; Hwang et al., 2020; Wang et al., 2023). These tools promote academic success and identify and evaluate knowledge gaps among learners, so it is important that educational and technological stakeholders, particularly policymakers and teachers, be well-versed in the potential and importance of AI for educational enhancement (Hwang et al., 2020; Kuddus, 2022).

The development of AI-based tools promises to provide substantial learning assistance for learners and teachers (Colak et al., 2008; Mohamed, 2024). Indeed, recent research indicates that AI-driven language learning tools offer several key advantages, including continuous accessibility, personalized learning experiences, and immediate feedback (Huang et al., 2023; Kuleto et al., 2021; Xie et al., 2019). Furthermore, research has specifically highlighted the role of AI in facilitating crucial aspects of language acquisition, such as student assessment and pronunciation enhancement (Pokrivcakova, 2019; Tlili et al., 2021). In addition to these benefits, AI-powered language learning programs can create engaging learning environments that effectively improve speaking skills (Jia et al., 2022). Moreover, AI-based technologies, such as chatbots, offer opportunities for English language practice at any time and location (Fathi et al., 2024; Fryer et al., 2020; Huang et al., 2022).

Chatbots, encompassing virtual assistants, teaching agents, and conversational agents, are versatile tools employed across diverse platforms and devices, including applications and websites (Bailey, 2019; Melian-Gonzalez et al., 2021). Advancements in chatbot technology, particularly the development of increasingly realistic features, have significantly impacted learner perceptions of their effectiveness in educational and non-educational settings. This evolving perception necessitates further investigation. In addition, research on chatbot integration has demonstrated improvements in several key areas, such as learning enjoyment, efficiency, and the social-emotional dimensions influencing English language acquisition (Golossenko et al., 2020; Pelau et al., 2021). However, while chatbots have been adopted across various domains, their specific application within English language learning remains relatively recent (Dale, 2016; Fryer et al., 2020).

Nevertheless, chatbots offer many advantages for both language learners and educators. These benefits include facilitating engaging conversations through voice and text, providing intelligent responses, promoting active interaction, enabling pronunciation practice, and delivering feedback. Additionally, these features contribute to improved learning outcomes, enhanced convenience, greater learner independence, increased comfort, and heightened confidence in the learning process (Huang et al., 2022; Kim, 2016; Walker & White, 2013). This potential has led second-language teachers and researchers to view chatbots as valuable conversational partners for English language learning (Jeon, 2022). Notwithstanding their supportive capabilities in delivering prompt responses and corrections, it is crucial to acknowledge a potential drawback. For example, students who rely too much on
AI technology may find it challenging to develop their critical thinking skills and independent thinking ability, correctly assess language, and understand linguistic patterns (Darwin et al., 2023).

Empirical research demonstrates that AI offers several benefits for language learners. Studies indicate that AI can foster students' emotional development, knowledge acquisition, and language learning progress (Rad, 2024). Furthermore, AI tools increasingly alleviate the workload associated with written assignments for students and educators by providing immediate feedback and accelerating the development of learners' writing skills (Marzuki
et al., 2023). In oral communication, AI assists learners in practicing and refining pronunciation and intonation while mitigating anxiety (Zou et al., 2023). Specifically,
AI-driven speech assessment systems have been shown to enhance learner motivation and confidence (Evers & Chen, 2020; Wang et al., 2022), thereby strengthening speaking proficiency in English language acquisition (Bashori et al., 2021; Kuddus, 2022; Tai & Chen, 2020).

However, research indicates that while English language teachers generally welcome integrating new technologies into their classrooms, they also face significant challenges
(An et al., 2021; Pokrivcakova, 2019). These challenges encompass practical concerns such as limited resources, inadequate training, conflicting attitudes toward technology, and the fears of professional displacement. Moreover, ethical considerations surrounding AI-based language learning systems are paramount (Mohamed, 2024). A key ethical concern centers on user data privacy (Wang et al., 2022). Furthermore, standardized AI platforms often overlook the nuanced diversity and sensitivity of cultural and human emotions (Jeon, 2022; Kim, 2017). While AI offers valuable enhancements to the learning process, it cannot fully substitute human interaction. Consequently, a balanced approach integrating AI and traditional teaching methodologies is crucial for an optimal learning environment (Mohamed, 2024; Pradana, 2023). Beyond ethical considerations, other challenges associated with technology integration in English language teaching include a lack of real-world application opportunities and technical and infrastructural limitations (Austin et al., 2017; Chien et al., 2020). Despite the promising potential of AI in language education, its successful implementation hinges on teacher satisfaction and the development of innovative curricula (Geng et al., 2021; Tsai & Chai, 2012). Therefore, providing teachers with comprehensive training that addresses both technical skills and ethical considerations is essential for effective AI integration in language education (Wang et al., 2022). Moreover, teacher adaptability, expertise, and proficiency in AI tools are critical for successful integration into English language teaching (Du & Gao, 2022; Pokrivcakova, 2019). Consequently, as AI continues to transform English language learning, addressing teachers' knowledge and skill gaps has become imperative to maximize the benefits of this technology in language education (Fakher Ajabshir, 2023).

The integration of AI into English language teaching and learning has garnered increasing attention due to its innovative potential and widespread accessibility. AI offers transformative opportunities to enhance language acquisition, retention, and learner engagement, challenging traditional pedagogical paradigms. Despite this promise, a comprehensive understanding of its effectiveness and practical application in English language learning (ELL) remains limited. Existing research often lacks cohesion, frequently emphasizing isolated case studies over systematic explorations of how AI can be seamlessly embedded into curricula. Qualitative research synthesis can bridge this gap by consolidating findings, identifying recurring themes and challenges, and establishing a foundation for robust pedagogical frameworks. While the benefits of AI-driven English language instruction are evident, addressing this research gap is critical, particularly given the broader implications for educational practice. A rigorous investigation into AI's applications, challenges, and best practices in this domain is thus essential to elucidate its capabilities thoroughly. Accordingly, this study undertakes a research synthesis to explore extant studies on AI in English language education, providing educators with insights and actionable recommendations for integrating this technology effectively. Drawing on a comprehensive analysis of 24 peer-reviewed articles from high-impact international journals, this study examines the impact of AI-driven pedagogy on English language acquisition and learner proficiency, with a particular focus on EFL instruction. Specifically, it addresses four research questions:

1. How is AI integration in English language teaching conceptualized?
2. What AI-based pedagogical approaches are most prominent?
3. Which variables are consistently associated with AI-mediated teaching?
4. What recurring challenges arise in its implementation?

Methods

This study undertakes a qualitative research synthesis to address the dispersed nature of existing research on AI in English Language Teaching (ELT). Rather than generating new data, this research re-analyzes and evaluates data from previously conducted studies (Cronin, 2011). Consequently, by integrating findings from diverse research endeavors, this review aims to provide novel insights and deepen our understanding of this subject (Sandelowski & Barroso, 2007). Specifically, this study analyzes existing research concerning the application of AI in ELT. To this end, a comprehensive search was performed across major international databases to identify relevant articles published between 2021 and 2024. The search employed "AI and EFL" and "AI in English Learning." This search yielded an initial pool of 158 articles, of which 24 met the established inclusion criteria and were subsequently selected for detailed review. In order to undertake a thorough assessment of the chosen papers, the research followed the seven-phase synthesis framework proposed by Sandelowski and Barroso (2007). The article selection process used the Critical Appraisal Skills Programme (CASP).

Figure 2. The Seven Stages of Synthesis (Sandelowski & Barroso, 2007, p. 105)

Step 1: Setting Research Questions

This study seeks to advance innovative methodologies in English language learning through empirical investigation and the provision of actionable insights. Specifically, this qualitative research synthesis examines AI-based EFL instruction with a threefold purpose. Firstly, it aims to identify and critically evaluate AI-based methods supported by empirical evidence demonstrating their efficacy. Secondly, it offers evidence-informed guidance on the pedagogical integration of AI technologies within English language classrooms. Finally, this study addresses crucial knowledge gaps in the field by delineating the most effective AI-based EFL instruction strategies. To achieve these objectives, this research analyzes the complex landscape of AI in EFL instruction and offers compelling recommendations for its optimal implementation.

Step 2: Systematic Review of Literature

A systematic literature review was conducted using a predefined set of keywords, as outlined in Figure 3, to explore the application of AI in English language teaching. This study analyzes selected peer-reviewed articles published in high-impact international journals, all written in English. The review incorporated a diverse selection of peer-reviewed English-language journals representing various geographical regions to examine the existing research thoroughly. Through this process, 24 relevant articles were identified and subjected to detailed analysis. These selected articles were retrieved from leading academic databases, including Scopus, Web of Science (WOS), ScienceDirect, SpringerLink, Wiley Online Library, Taylor & Francis Online, and Sage Journals.

Step 3: Search and Select

The third phase of the current synthesis involved a rigorous appraisal of the quality of studies selected from multiple databases. This quality assessment process adhered to a four-stage protocol. Initially, titles, abstracts, and available content were screened. Subsequently, a comprehensive quality assessment was conducted using the Critical Appraisal Skills Programme (CASP) tool, specifically designed for evaluating primary qualitative research. This evaluation involved a detailed review of each study's research objectives, methodology, design, and sampling strategy. Each article's title, abstract, methodology, discussion, and conclusion were meticulously examined to ensure adherence to predefined inclusion criteria, such as thematic relevance and full-text availability. A standardized
ten-item CASP checklist, with a scoring range of 0 to 50, was employed to categorize articles into five quality tiers: excellent (41-50), perfect (31-40), good (21-30), moderate (11-20), and poor (0-10). Consequently, only articles achieving a good or excellent rating were advanced to subsequent review stages. This stringent quality control excluded 134 of 158 appraised articles, leaving 24 eligible for an in-depth analysis. This was followed by the establishment of an inter-coder agreement to guarantee uniformity in the selection and rating process (Cohen's kappa, resulting in a score of 0.88 (p < 0.05)), wherein both researchers separately evaluated the studies according to pre-defined criteria that the researchers had already developed (Table 1). The article search and selection process are visually represented in a flowchart (Figure 3).

Table 1. Inclusion and Exclusion Criteria for Articles in the Synthesis Process

Figure 3. Flowchart of the Inclusion of Articles in the Synthesis Process

Step 4: Extraction of Articles Information

A comprehensive data extraction table (Table 2) was developed to encapsulate the salient features of the studies included in this review. This table includes the research’s title, authors, publication year, study location, research methodology, sample size, and measurement instruments. Furthermore, to facilitate subsequent data analysis and coding procedures, each article underwent a meticulous review, during which pertinent information relevant to the research questions was extracted and documented.

Table 2. Characterization of Included Studies

Analysis of articles indexed in scientific journals from 2021 to 2024 reveals a significant increase in research concerning the application of AI in English instruction, with a particularly notable surge occurring in 2023 (Figure 4). This growth in research output is further contextualized by a geographical analysis, which indicates a concentration of research activity originating from China. Moreover, South Korea and Iran have also emerged as prominent contributors to the expanding literature on AI in English language education (Figure 5). Figure 4 visually depicts the upward trend in publication numbers over the specified period; Figure 5, on the other hand, offers a thorough summary of the worldwide distribution of research activity according to the locations of data and study samples that have been gathered. It highlights significant regional centers and illustrates the global reach of interest in AI-based language learning.

Figure 4. Line Graph of the Number of Indexed Articles per Year

Figure 5. Bar Chart of the Geographical Distribution of Research

Note: Based on the Collected Data and Study Samples' Locations

Step 5: Analysis and Synthesis of Results

A qualitative evaluation of the data compiled in Table 3 was performed. With the aid of MAXQDA 24 software, pertinent themes were drawn from the findings and discussion portions of the chosen research papers. This analysis led to the development of a structured coding system, starting with 42 basic open codes, which were then grouped into 14 intermediate axial codes and then condensed into five broad selective codes (categories).
A second researcher (Second Author) separately analyzed a randomly picked portion of the data to verify consistency between coders. The level of agreement between the two coders was measured using Cohen's kappa, resulting in a value of 0.81—a statistically meaningful figure (p ≤ 0.05)—demonstrating strong alignment. This thorough approach reduced personal bias and bolstered the trustworthiness of the derived categories. The entire coding structure, complete with explanations and supporting examples, is outlined in Table 3, ensuring clarity and reinforcing the credibility of the research outcomes.

Table 3. Categorization of Open, Axial, and Selective Codes (Themes)

Step 6: Quality Control

Several critical factors were meticulously addressed to ensure the highest quality standards in this research. First, a strong emphasis on transparency was maintained throughout all stages of the study, encompassing the careful selection of articles and a thorough analysis of their findings. Second, an exhaustive review of resources relevant to integrating AI into EFL education was conducted. This comprehensive literature review spanned various periods, enabling a robust and validated comparison of results. Furthermore, Cohen's kappa coefficient was employed to establish the reliability and validity of the findings. Subsequently, two expert university faculty members specializing in educational technology within language learning meticulously evaluated the results. Both experts agreed on extracting distinct concepts, axes, and themes. The kappa index was again applied further to validate the quality of these extracted concepts and categories. The resulting kappa coefficient for this research was calculated to be 0.86, a statistically significant value at (p ≤ 0.05). Therefore, it can be confidently concluded that the extracted codes demonstrate high reliability.

Step 7: Presentation of Findings

This research synthesis's seventh and final phase comprehensively integrates the findings from all preceding stages. Specifically, through analysis of the assembled research, this culminating phase directly addresses the central research questions concerning the conceptualization of AI integration within English language learning. This analysis encompasses a detailed examination of the included studies' methodological foundations, identifying key variables influencing AI integration and critically evaluating the challenges encountered during AI implementation. Consequently, these synthesized elements are visually represented and theoretically grounded within a comprehensive conceptual model (Figure 6).

Figure 6. Conceptual Model Drawn from Research Results

Table 3 presents the coding results derived from the research findings, identifying five principal themes and 21 subthemes. These themes encompass a range of perspectives on AI in education. Firstly, the "Conceptualization of AI" theme explores simplifying the learning environment, fostering collaboration with technology, and enhancing overall learning outcomes. Secondly, "Variables related to AI" examines the impact of AI on various aspects of the learning process, including providing emotional support, increasing participation and social interaction, promoting critical thinking and creativity, delivering timely feedback, improving language learner performance, creating effective learning environments, enhancing motivation, and addressing psychological factors. Thirdly, "Integration challenges" consider the practical aspects of implementing AI in educational settings, focusing on usability, enjoyment, recency effects, and the complexities of technological integration. Fourthly, "Limitations" addresses the inherent constraints of
AI in education, such as technological limitations, specific issues and obstacles encountered, and the current lack of personalization. Finally, "The Future of AI (Foresight)" explores potential developments, including the current initial stage of AI integration, concerns regarding privacy protection, the pursuit of educational justice, and the recognized need for further research and development in this field. The analysis encompassed 24 studies, with a majority conducted within Asia. Notably, seven studies originated from China. While variations were observed across the studies, likely attributable to differences in sample populations and methodologies, the qualitative data generally pointed towards positive outcomes of AI in the EFL curriculum.

Furthermore, Figure 7 illustrates the thematic density of the extracted themes. This visualization reveals a primary emphasis on "Variables related to AI in English learning," suggesting a strong focus on AI's direct impacts on learning. Conversely, "Limitations" and "Integration challenges" received comparatively less attention, highlighting the need for further research and investigation in these areas.

Figure 7. Radar Chart of Qualitative Analysis

Discussion and Conclusion

AI has revolutionized English language education by providing individualized, interactive learning experiences and helping instructors with curriculum design, grading, and progress tracking, increasing efficiency, effectiveness, and production (Mohamed, 2024). This study interpreted comprehensive findings from prior research on AI's impact on English language learning.

Initially, research indicates that AI-driven tools enrich learning beyond the classroom, simplify teaching methodologies, and improve learners' academic performance (Mohamed, 2024; Qasem et al., 2023). Researchers generally agree that AI in English language learning enhances access to information, fosters creativity and critical thinking (Huang et al., 2021; Wang & Kokotsaki, 2018), facilitates more straightforward assessment and evaluation, promotes learner collaboration, and reduces learning isolation (Ghafouri, 2024).

Furthermore, integrating AI into English language education supports self-directed learning, offering numerous benefits to language learners (Darwin et al., 2023; Kartal, 2024; Hong, 2023). Specifically, AI-powered interactive speaking practice has significantly enhanced learners' speaking abilities and motivation (Fathi et al., 2024; Vladova et al., 2021). These programs also provide personalized feedback, boosting learner confidence and contributing to positive academic outcomes (Rad, 2024) while effectively addressing learners' emotional needs (Wong et al., 2022). Consequently, by customizing experiences, providing immediate feedback, and creating supportive environments, AI-powered tools demonstrably enhance English language acquisition, explicitly improving speaking skills, bolstering confidence, and increasing willingness to communicate.

Moreover, researchers suggest that AI can enhance English language learning by personalizing instruction, strengthening teacher-learner and learner-learner collaboration, reducing stress, and providing practical solutions to teaching obstacles (Ghafouri, 2024). Various AI-powered tools enrich the learning experience by improving learners' writing abilities and creating a more comprehensive learning environment (C. Liu et al., 2021). Beyond simply enhancing scores, AI writing tools offer promising capabilities for learners to effectively express ideas, enrich content, facilitate idea generation, and creatively solve problems (Marzuki et al., 2023; Y. Liu et al., 2022). While these tools enhance academic achievement and positively impact writing coherence and structure, concerns regarding over-reliance remain (Marzuki et al., 2023). Nonetheless, integrating AI writing tools can ultimately elevate learners' writing quality. Regarding speaking skills, AI programs offer ample practice and improve English language speaking through interactive and adaptive methods (Chang et al., 2022; Rad, 2024). AI speech assistants provide personalized experiences, help adjust accents and control speaking speed, foster comfortable interactions, and enhance language skills, confidence, and enthusiasm (Zhang et al., 2024). Finally,
AI chatbots reduce stress and increase comfort in practicing speaking in stress-free environments, improving fluency, coherence, vocabulary, grammar, and pronunciation (Chaisiri, 2023; Fathi et al., 2024; Ma et al., 2022).

Beyond skill development, AI technology offers numerous psychological and emotional benefits in English language learning. Furthermore, AI can strengthen the teacher-learner relationship, reduce negative emotions, promote learner interaction, and thus increase the sustainability of English language learning (Ghafouri, 2024; Xie & Derakhshan, 2021), contributing to learners' enjoyment (Zhang et al., 2024). Although chatbots have limitations in interaction and are prone to errors, they aid language learning by improving skills and reducing anxiety (Alrajhi, 2024). Learners view chatbots as valuable tools for effective interaction. Also, teacher perceptions influence classroom use (Divekar et al., 2021; Fryer et al., 2020; Jeon, 2022; Timpe-Laughlin et al., 2020). Beyond enjoyment and efficiency,
AI tools significantly impact learners' social and emotional perceptions (Kwon et al., 2023; Pelau et al., 2021), providing opportunities for conversation practice and improving various aspects of language learning, including skills and motivation (Ebadi & Amini, 2022; Han & Lee, 2022). Consequently, AI literacy, enabling teachers to evaluate, communicate with, and utilize this technology, becomes crucial (Liu et al., 2024; Weng & Chiu, 2023). Chatbots that mimic human interaction can increase learner motivation and social presence (Araujo, 2018), and integrating AI communication tools can enhance motivation and learning experiences in English classrooms (Araujo, 2018; Ebadi & Amini, 2022).

However, while acknowledging AI's benefits and growing potential in English language education, it is essential to recognize its limitations and challenges. Research emphasizes the need for accurate evaluation and understanding of AI tool limitations before its integration to maximize benefits and minimize drawbacks (Rudolph et al., 2023). Researchers have stressed the need for conscious use of AI technology due to its shortcomings (Cotton et al., 2023), identifying challenges such as inaccurate responses, reduced learner independence (Dwivedi et al., 2023), and limitations in accuracy and cultural sensitivity (Zhao, 2023). Effective integration requires assessing learners' needs (Mohamed, 2024). Despite their widespread use, Chatbots face challenges in understanding mistakes and replicating human experiences (Fathi et al., 2024), while worries about excessive dependence continue (Kasneci et al., 2023). Given the importance of teacher-student relationships, AI should complement traditional methods, preserving the teacher's central role (Mohamed, 2024). To optimize AI's benefits, teachers require technical and specialized training before integrating these tools (Chen et al., 2020; Holmes & Tuomi, 2022), as this has shown positive impacts on teaching effectiveness, skills development, practice opportunities, and classroom interaction. Based on these limitations, effective AI integration requires empowering learners with independent learning methods, targeted assignments, and a proper understanding of AI's impact. Responsible integration also necessitates educating learners about data collection and other related concerns. Therefore, AI tools can enhance English language education through conscious, planned, balanced integration that maintains learner motivation.

Finally, integrating AI tools into face-to-face instruction to facilitate information retrieval, provide targeted feedback, cultivate critical thinking, and personalize learning pathways offers another promising strategy. By tailoring experiences to individual needs and interests, AI has the potential to enhance classroom outcomes significantly. However, further research is crucial to exploring and fully optimizing these tools' pedagogical applications.

Limitations

This study acknowledges several limitations in its data collection and analysis. The review was confined to sources from 2021 to 2024, excluding relevant pre-2021 studies, which may impact the comprehensiveness. The focus on peer-reviewed scholarly articles from reputable databases also overlooked insights from alternative sources like organizational reports. The research synthesis methodology, while beneficial, remains subject to the researcher's biases despite efforts to uphold methodological rigor. The findings, drawn from various educational systems, cannot be generalized without considering each context's unique characteristics. Furthermore, the variability in statistical samples across the articles posed challenges in comparison and synthesis. Despite these limitations, the study aims to provide a comprehensive overview of the benefits and challenges of integrating AI into EFL classrooms. We hope this study will catalyze further research in this area, encouraging scholars to explore the potential of technology-based language instruction.

Acknowledgments

The authors express their sincere gratitude to all who supported this research.

Declaration of Conflicting Interests

The authors disclaim any potential conflicts of interest regarding the research, writing, and/or publishing of this work.

Funding

The writers did not receive any funding to conduct this study, writing, and/or publishing.