The integration of AI-mediated instruction in EFL education requires a nuanced understanding of how motivation and behavior shape learner engagement. Central to this discussion is Self-Determination Theory (SDT) by Deci and Ryan (1985, 2000) and Ryan and Deci (2020), which remains a key framework for examining motivation in educational settings. According to SDT, motivation lies on a continuum from external to internal regulation and is influenced by the satisfaction of three basic psychological needs: autonomy, competence, and relatedness. Recent studies show that AI-enhanced environments can affect these motivational factors in English language learning (Ali et al., 2023; Aly et al., 2024; Ebadi and Amini, 2024; Fandiño et al., 2019; Hao et al., 2024; He, 2024; Madinabonu, 2024; Moybeka et al., 2023; Ramadhani, 2025; Silitonga et al., 2023; Suciati et al., 2024; Tran, 2024; Wang F. et al., 2025; Yang T., 2024; Yaşar and Karagücük, 2024).

However, frequent interaction with AI tools does not necessarily reflect internal motivation. To explore this further, Fredricks et al.’s (2004) tripartite model of engagement—covering behavioral, emotional, and cognitive dimensions—offers a valuable lens. While behavioral signs like task completion are easy to track, they may not reflect deeper learning or motivation, especially when AI tools lead to surface-level engagement. Although studies have shown AI’s positive effect on behavioral outcomes (Bhatt and Muduli, 2024; Ellikkal and Rajamohan, 2024), other findings suggest that overreliance on GenAI may weaken critical thinking and personal effort (Alasadi and Baiz, 2023; Pido et al., 2025). Waluyo and Kusumastuti (2024) also warn that high engagement with GenAI does not always lead to better academic results, further challenging the link between AI use and authentic learning gains.

This complexity calls for a more careful look at how GenAI shapes learners’ motivation and engagement. Reeve and Tseng (2011) and Skinner (2016) emphasize that externally driven behavior should not be mistaken for genuine commitment. In AI-supported EFL contexts, visible engagement may not indicate internal motivation, showing that behavioral data alone may be insufficient for understanding true learning progress.

To address this paradoxical phenomenon, the present study introduces the construct of the “Engaged but Amotivated” (EBA) learner. Drawing on Self-Determination Theory (Deci and Ryan, 1985, 2000; Ryan and Deci, 2020), EBA refers to students who actively participate in learning tasks—often driven by AI prompts or institutional expectations—yet report low intrinsic motivation, limited emotional connection, and even low academic performance. Specifically, EBA students demonstrate consistent behavioral participation—such as task completion, punctual submission, and frequent AI interaction—without exhibiting corresponding intrinsic or identified motivational regulation. In short, these learners are “doing” without genuinely “wanting.”

In AI-mediated environments, this disconnect between action and motivation is often obscured by the high visibility of behavioral participation. The gamification of tasks, automated feedback loops, and performance-tracking mechanisms commonly integrated in GenAI platforms may incentivize learners to complete tasks efficiently, but not necessarily meaningfully. According to Fredricks et al. (2004), true engagement involves not only behavioral indicators but also cognitive and emotional involvement. In the case of EBA learners, these latter dimensions often remain underdeveloped, with students demonstrating a passive orientation to tasks and limited evidence of deeper cognitive processing. Additionally, insights from cognitive load theory (Sweller, 2011) offer a complementary perspective. When low-proficiency learners overly depend on AI to generate content, their active involvement in the high intrinsic load and the overall cognitive process essential for language development may be diminished.

While the EBA construct is new, it overlaps with—but is also distinct from—existing ideas such as amotivation, surface learning, and performance-approach learning. Amotivation means a student has no motivation and usually does not take part in learning activities (Deci and Ryan, 1985, 2000; Ryan and Deci, 2020). In contrast, EBA students appear active and complete tasks but feel no real interest or purpose. EBA is harder to notice because the lack of motivation is hidden by visible participation. Surface learning refers to the learning with “motive of meet institutional requirements minimally, and the congruent strategy is limiting the target to essentials that may be reproduced through rote learning” (Biggs, 1988, p. 129). Although surface learning and EBA both show limited cognitive engagement, they differ in motivation and behavior. Surface learners are extrinsically motivated and focus on minimum requirements. In contrast, EBA learners appear highly active but lack meaningful motivation. Performance-approach learning emphasizes outcome-based goals (Elliot, 1999; Elliot and Moller, 2003; Elliot et al., 2005; Gilbert and Elliot, 2024) where learners are driven by grades or comparison, which can still involve high motivation, albeit extrinsic. In contrast, EBA uniquely combines high visible effort with an inner motivational void. It describes students who appear engaged but experience emotional detachment and a lack of personal meaning in their learning actions.

Recent studies support this pattern. Liu et al. (2024a) found that Chinese university learners using AI tools in informal digital learning often engage out of habit or utility rather than personal interest, particularly when autonomy is limited. Similarly, Pan and Wang (2025) and Zhou and Hou (2024) noted how AI’s efficiency and structure can undermine the human elements of teaching and learning, raising concerns about over-reliance and emotional disconnection. Shen et al. (2024) further argued that emotional support from teachers remains essential for fostering motivation and further engagement, suggesting that AI, by itself, cannot sustain learner motivational well-being.

In sum, the EBA construct draws attention to a hidden problem in AI-mediated education: the illusion of engagement. It reminds educators and designers that participation does not always equal meaningful learning. Clarifying this concept helps separate active behavior from authentic motivation, which is essential for responsible use of AI tools in EFL contexts.

Chinese HVCs represent a unique and often overlooked setting within the broader context of GenAI-enhanced EFL education. Enrolling over half of the nation’s tertiary students (Ministry of Education of the People’s Republic of China, 2024; Xu, 2025), these institutions primarily serve learners who are placed into vocational tracks through academic streaming systems such as the gaokao (高考) (Xiong, 2011). As a result, students in HVCs often begin their English studies with lower academic readiness, weaker language proficiency, and a more practical orientation toward language learning (Pan and Dapat, 2023; Wang, 2024; Wang and Chen, 2021). These characteristics position HVCs as a key context for understanding how GenAI affects motivation, learning behavior, and learner identity differently than in traditional university settings (Wang, 2024; Wei W. et al., 2025).

Despite national goals for inclusive digital transformation, HVC learners are frequently absent from both policy and empirical discussions. Most studies of AI-based EFL instruction in China focus on university students with stronger self-regulation, academic ambition, and digital literacy (Liu et al., 2025; Xu and Liu, 2025). In contrast, vocational learners may struggle with the self-directed learning and critical evaluation skills needed to use GenAI tools effectively. This makes them particularly susceptible to the EBA pattern, where task completion masks limited internal engagement.

The institutional environment adds further complexity. Curricula in HVCs are closely aligned with vocational training and often treat English as a skill to pass tests or secure employment rather than as a tool for personal or intellectual growth (Cao and Chen, 2023; Wei Q. et al., 2025; Wei W. et al., 2025). Within this framework, AI systems that reward efficiency may unintentionally encourage mechanical completion of tasks, rather than deeper processing or genuine interest. Students may depend on AI-generated content or aim for performance metrics without actively engaging in language development.

Sociocultural influences further shape learners’ attitudes. Many HVC students come from lower-income families and face pressure to prioritize immediate employability. Traditional values such as diligence and conformity (Cheng, 2020) may align with AI’s output-driven design, reinforcing behaviors aimed at meeting visible expectations rather than fostering reflective or self-directed learning. In this context, GenAI may be viewed as a functional but impersonal tool, especially if it challenges learners’ confidence or sense of relevance (Riser, 2025).

Given these dynamics, it is essential to examine GenAI use in HVCs through a nuanced, learner-centered lens. The question is not simply how to integrate technology, but how to align it with learners’ needs and realities. Without thoughtful pedagogical support, GenAI risks reinforcing superficial learning and widening existing gaps. Constructs like EBA can help identify misalignments between participation and meaningful engagement. A context-sensitive approach that fosters learner autonomy and emotional connection is key to ensuring that technology acts as an enabler rather than a barrier.

The study was conducted at a first-year class at an HVC located in eastern China. Firstly, located in the Yangtze River Delta, a region recognized as the most economically and technologically advanced in China, this college was purposefully selected due to its proactive incorporation of GenAI tools into first-year EFL classes as part of a broader digital education reform initiative. Participants were selected through purposive sampling, targeting students actively engaged in GenAI-mediated English classes but exhibiting varied motivational profiles. The final sample included 39 first-year students majoring in Digital Media, chosen for their consistent exposure to GenAI tools—such as Xuexitong, Doubao(豆包), DeepSeek, and Kimi—in class and for assignments (Figure 1).

The participant group was predominantly female and aged 19, reflecting the typical profile of first-year students in vocational EFL programs. Most students (89.74%) reported English proficiency at or below CET-3, which roughly corresponds to the A2 to low B1 levels on the Common European Framework of Reference for Languages (CEFR). This indicates limited language competence, with learners capable of basic communication but often struggling with academic English—likely influencing their reliance on GenAI tools and contributing to patterns of extrinsically driven engagement. Nearly all participants had prior experience using GenAI tools in class, with over 60% using them for more than 6 months. A majority (76.9%) were from Jiangsu Province, providing a relatively consistent regional context.

Data collection spanned two academic semesters (approximately 36 weeks), providing longitudinal insights into learners’ engagement and motivational dynamics within consistent instructional settings.

To obtain a comprehensive understanding of the participants’ experiences with AI-mediated English instruction, qualitative data from interviews, classroom observations, and LMS log records were subjected to thematic analysis. This approach allowed for the systematic identification and interpretation of patterns within the data, offering rich insights into the EBA tension.

A central theme identified in participant accounts is the notion of performative engagement, where learners engage with GenAI tools—such as DeepSeek, Doubao (豆包), and Kimi—primarily to fulfill institutional requirements rather than to pursue meaningful learning. These tools are commonly used for routine academic tasks such as grammar correction, sentence translation, and writing assignments. While such use may appear to demonstrate active participation, especially through analytics on learning platforms, closer analysis reveals a surface-level engagement driven by external demands.

When asked, “Do you learn English because you are interested, or mainly because it’s required?,” only 6 out of 39 participants cited personal interest or communicative purposes. One remarked, “I’m personally interested in English” (Appendix 1–Q1–1), while another shared, “I have some interest; I want to travel abroad and communicate with foreigners” (Appendix 1–Q1–2). In contrast, the remaining 33 students indicated extrinsic motivations, linking their learning to institutional goals such as passing exams, gaining credits, or meeting course requirements. As one student noted, “It’s mainly a school requirement; I’m mostly passive and grade-focused” (Appendix 1–Q1–3), while another explained, “The main reason is school requirements—completing assignments and passing tests to earn credits” (Appendix 1–Q1–4). Additionally, many students reported learning English primarily to prepare for the vocational-to-university transfer examination. As one student shared, “I mainly study English for the transfer exam” (Appendix 1–Q1–5), while another echoed, “For the transfer, I must pass English” (Appendix 1–Q1–6). These responses further illustrate that students’ engagement with English is largely driven by institutional and exam-related demands rather than by personal interest or intrinsic motivation.

These extrinsic drivers are reflected in classroom behavior. When asked, “How do you behave during AI-assisted English classes? Are you active or passive?,” 25 out of 39 students described themselves as “passive” or “somewhat passive” (Appendix 1–Q2–1), suggesting a lack of personal agency or enthusiasm. Moreover, in response to the question, “Do you feel more like a learner or a tool user when working with AI?,” only 6 participants identified as learners; the rest viewed themselves more as tool users. One participant reflected, “I feel more like a tool user and do not take initiative in learning” (Appendix 1–Q3–1).

These reflections reveal a mismatch between external engagement and internal motivation. While institutional systems may read tool usage as active learning, many students describe their actions as a form of obligation rather than authentic learning. In this sense, GenAI tools become instrumental means to complete academic tasks, often replacing the cognitive effort and intrinsic interest that meaningful learning requires.

Many participants reported a noticeable decline in their motivation to learn English during college. This reduction was frequently linked to increased academic pressure, the complexity of learning tasks, and persistent struggles with vocabulary and grammar. Although some students acknowledged the importance of English for exams or future job opportunities, these instrumental goals were often overshadowed by emotional fatigue and anxiety—particularly around high-stakes exams such as the CET-4. Several learners also noted that English felt increasingly disconnected from their personal interests or everyday needs.

Compared to their earlier schooling, where teacher support and structured classroom routines occasionally sparked interest, the transition to more self-directed college learning was often described as demotivating. Without sufficient guidance, autonomy led some students to disengage. For example, one remarked, “Now English is just a task to complete—I do not feel motivated” (Appendix 2–Q4–1). While a few did experience satisfaction when successfully using GenAI tools, such moments were rare and insufficient to sustain meaningful motivation.

In response to the question, “Has your motivation for learning English changed since using AI tools?,” only 8 out of 39 participants reported increased motivation. One student explained, “Yes, I feel that learning English is not as difficult anymore” (Appendix 2–Q4–2). However, 12 students described growing dependence on AI tools rather than increased engagement. As one stated, “I feel increasingly dependent because AI is too powerful and limits students’ thinking” (Appendix 2–Q4–3). Thirteen participants reported “no change” (Appendix 2–Q4–4) in their motivation, while a few mentioned feeling “less motivated” (Appendix 2–Q4–5). One noted, “After using AI for a while, I do not really want to think by myself” (Appendix 2–Q4–6).

These reflections suggest that although GenAI is perceived as useful for academic tasks—especially grammar correction and translation—it does not consistently lead to deeper or sustained motivation. One contributing factor appears to be cognitive overload. For students with lower English proficiency, using AI required additional mental effort—choosing tools, understanding complex outputs, and navigating digital interfaces. Rather than simplifying the learning process, this sometimes made it more overwhelming.

This burden was mirrored in students’ emotional responses. While a few described curiosity or brief satisfaction, many reported emotional detachment, using expressions such as “indifferent,” “no feeling,” “calm,” or “bored” (Appendix 2–Q4–7). Some also mentioned anxiety or confusion, especially when uncertain about how to use AI effectively. Taken together, these responses highlight that for many learners—especially those with limited digital or language confidence—GenAI tools are seen more as convenient aids than as meaningful sources of engagement or motivation.

Participants’ reflections revealed a dual perspective on the role of GenAI in shaping their identities as English language learners. On the one hand, many students regarded AI tools as empowering supports that reduced anxiety and increased confidence. Several described becoming more willing to engage in learning tasks. One student shared, “I’ve become more proactive” (Appendix 3–Q5–1), while another remarked, “It made me more confident and active” (Appendix 3–Q5–2). For these learners, GenAI served as a cognitive scaffold that helped bridge linguistic gaps and supported independent learning, particularly in reading, writing, and comprehension. As a result, GenAI was seen as an enabler that made English learning feel more accessible and less intimidating.

On the other hand, a comparable group of students expressed concerns that overreliance on GenAI was weakening their sense of ownership and engagement. Some reported reduced motivation and a lack of challenge. One participant stated, “With AI, learning English no longer feels challenging” (Appendix 3–Q5–3). Others voiced stronger disengagement, as shown in the comment, “It’s boring; learning English feels useless” (Appendix 3–Q5–4). In contrast, several students emphasized the need for personal effort, stating, “Progress depends on oneself and the willingness to learn” (Appendix 3–Q5–5), and “Progress relies not only on AI, but also on self-discipline, hard work, and determination” (Appendix 3–Q5–6). These responses suggest that while GenAI can support learning, it may also discourage reflection and reduce students’ active participation in the learning process.

This ambivalence became more pronounced when participants were asked, “If AI tools were no longer used in your English class, how would you feel?” Many described strong negative emotions, including “anxiety,” “loss,” and “inconvenience.” One student commented, “I would feel anxious because without AI, I truly would not understand anything” (Appendix 3–Q6–1). Another explained, “I would feel lost and anxious, as if my work would no longer be complete or perfect” (Appendix 3–Q6–2). These statements highlight the deep integration of AI into their academic routines. At the same time, a smaller number of students welcomed the potential benefits of AI removal. One reflected, “I would feel a bit panicked, but it might help me think and work independently” (Appendix 3–Q6–3), and another shared, “It would encourage independent thinking and help me improve” (Appendix 3–Q6–4). A few students were indifferent, saying “no effect” or “it does not matter” (Appendix 3–Q6–5).

In summary, learners’ perspectives on GenAI reveal a complex interplay of dependence and agency. While many appreciate the confidence and convenience it offers, others caution against its potential to limit critical thinking and personal growth. This tension illustrates a broader educational challenge: integrating AI in ways that enhance, rather than replace, learners’ autonomy and identity development.