I advocate for AI in education loudly and unapologetically. But advocacy without nuance is reckless. And this study from Fan et al. (2025) is the kind of research that keeps my advocacy grounded.
Their paper, “Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance,” published in the British Journal of Educational Technology, introduces a concept I think every educator needs to sit with: metacognitive laziness. Straightforward but unsettling. When students use AI, they may stop monitoring their own thinking. They offload the cognitive work that actually builds understanding.
Fan et al. define metacognitive laziness as “learners’ dependence on AI assistance, offloading metacognitive load and less effectively associating responsible metacognitive processes with learning tasks.” (p. 506).
I’ve been writing about cognitive offloading and cognitive surrender for months now. Metacognitive laziness adds another layer. Students defer to AI outputs (Shaw & Nave, 2026). Heavy AI use weakens critical thinking (Gerlich, 2025). And now we have evidence that the self-regulation processes students need to learn effectively, orientation, monitoring, evaluation, get quietly displaced when AI handles the revision work.
The Experiment
Fan et al. conducted a randomized lab experiment with 117 university students. Participants completed a two-stage English reading and writing task. Four groups were compared: no support (control), ChatGPT 4.0 support, human expert support, and checklist-based writing analytics tools. Motivation, self-regulated learning processes, essay improvement, knowledge gain, and knowledge transfer were all measured.
What makes this design strong is the comparison structure. AI is tested against a human expert and a structured tool, not just a control. That gives us a much richer picture.
On essay improvement, the AI group significantly outperformed everyone else: “the AI group significantly improved the essay scores compared to other groups” (p. 507). ChatGPT was especially effective when a clear rubric guided the task. It optimized content around visible criteria, reorganized structure, and strengthened surface-level quality.
If your goal is a better essay draft, AI delivered. No question.
Two findings complicate that picture.
Motivation didn’t budge. There were “no significant differences in intrinsic motivation among the four groups” (p. 505). AI didn’t boost it. Didn’t hurt it either. But the widespread assumption that AI tools energize students? This study doesn’t support it.
Knowledge gain and transfer tell a deeper story. Fan et al. report “no significant differences in knowledge gain or transfer” (p. 507). Students in the AI group wrote better essays but didn’t learn more. They didn’t transfer knowledge to new contexts any better than the control group.
That’s a critical finding. AI improved the product without improving the process. The essay got better. The student didn’t.
Process Mining Reveals the Problem
Fan et al. used trace data to map how students regulated their learning during revision. And this is where the concept of metacognitive laziness comes to life in the data.
Students in the AI group formed tight loops between revising and interacting with ChatGPT. Bouncing back and forth, refining text based on AI suggestions. But they showed relatively fewer metacognitive processes (eg, evaluation and orientation) compared to the human expert and checklist groups (p. 506).
With a human expert, the pattern looked different. Stronger connections between revising, orientation, and evaluation. Pausing. Assessing. Planning. The expert didn’t just give feedback. The interaction itself prompted metacognitive engagement.
Kosmyna et al. (2025) found something parallel at MIT. ChatGPT use reduced neural engagement during writing tasks, but students who thought independently before using AI produced stronger outputs. AI streamlines the task. It may also thin out the cognitive activity that makes the task educational.
What This Means for AI Pedagogy
I want to be clear. Fan et al. don’t argue against using AI in learning. They frame their work within hybrid intelligence, defined as a “combination of human and machine intelligence, augmenting human intellect and capabilities instead of replacing them and achieving goals that were unreachable by either humans or machines” (Akata et al., 2020, p. 19 cited in Fan et al., 2025, p. 491). Their concern is about how AI is used, not whether it should be.
And that aligns with everything I’ve been arguing. Pedagogy determines whether AI supports or weakens learning. Guo et al. (2025) showed across a full year of classroom use that activity type and instructional design shaped ChatGPT’s educational value. Cheng et al. (2025) found that student agency in the AI interaction, asking purposeful questions, predicted writing improvement. Sperber et al.’s (2025) PAIRR model builds structured reflection after AI feedback to develop critical evaluation skills.
Fan et al. add an important piece to this puzzle. Even when AI improves task performance, educators need to watch for metacognitive laziness. Are students evaluating AI suggestions against their own understanding? Or just looping between “revise” and “ask ChatGPT” without ever pausing to think about why?
As the authors caution: “ChatGPT can quickly improve task performance, it does not significantly enhance intrinsic motivation and may trigger metacognitive laziness” (p. 507). Worth printing out and taping above every computer in a writing lab.
The Takeaway
AI can make student writing better. That’s clear from this study and many others. But better writing doesn’t automatically mean better learning. Metacognitive laziness lives in the gap between improved output and improved understanding. Intentional pedagogy needs to fill it.
Build in reflection. Require students to evaluate AI feedback before acting on it. Ask them to explain their revision decisions. The essay matters. What happened in the student’s head while writing it matters more.
Reference
- Akata, Z., Balliet, D., De Rijke, M., Dignum, F., Dignum, V., Eiben, G., Fokkens, A., Grossi, D., Hindriks, K., Hoos, H., Hung, H., Jonker, C., Monz, C., Neerincx, M., Oliehoek, F., Prakken, H., Schlobach, S., van der Gaag, L., van Harmelen, F., … Welling, M. (2020). A research agenda for hybrid intelligence: Augmenting human intellect with collaborative, adaptive, responsible, and explainable artificial intelligence. Computer, 53(8), 18–28.
- Cheng, Y., Fan, Y., Li, X., Chen, G., Gašević, D., & Swiecki, Z. (2025). Asking generative artificial intelligence the right questions improves writing performance. Computers and Education: Artificial Intelligence, 8, 100374. https://doi.org/10.1016/j.caeai.2025.100374
- Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X., & Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56(2), 489–530. https://doi.org/10.1111/bjet.13544
- Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), Article 6. https://doi.org/10.3390/soc15010006
- Guo, F., Li, T., & Cunningham, C. J. L. (2025). One year in the classroom with ChatGPT: Empirical insights and transformative impacts. Frontiers in Education, 10, 1574477. https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2025.1574477/full
- Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing tasks. MIT Media Lab. https://www.media.mit.edu/publications/your-brain-on-chatgpt/
- Sperber, L., MacArthur, M., Minnillo, S., Stillman, N., & Whithaus, C. (2025). Peer and AI Review + Reflection (PAIRR): A human-centered approach to formative assessment. Computers and Composition, 76, 102921. https://doi.org/10.1016/j.compcom.2025.102921