
Opinion
Students are increasingly seeking out GenAI tools to help with their studies, from understanding difficult readings to improving assignments. This is becoming increasingly easier, for example with many PDF readers now containing inbuilt AI chatbots which offer to summarise, translate, simplify and explain a text to students as soon as they open a class reading. Similarly, AI tools are increasingly built into word processors, suggesting improvements to an assignment and offering reassurance to a worried student. Unlike standalone tools such as ChatGPT and Claude, which students have had to seek out, these inbuilt tools now require effort to avoid.
When these systems choose what deserves attention, what is relevant in a reading and what needs improving in an assessment, they gain influence over students’ education. Arguably, given the frequency with which some students use AI in their educational studies these systems may potentially have a greater influence than their own lecturers.
Understanding when GenAI supports learning and when it displaces it requires continued research, both conceptual and empirical. This includes listening to students about how and why they use it. The AIinHE project (aiinhe.org) surveyed students at four Australian universities (Monash, UQ, UTS and Deakin), involving more than 8,000 students in 2024 and more than 10,000 in 2026. Their responses offer insight into AI’s appeal and how students’ reported use and attitudes have changed between the two surveys.
As you might expect, student use of GenAI for study purposes is increasing. Weekly or daily GenAI use for study purposes increased from 44% in 2024 to 55% in 2026. Students’ engagement with GenAI was strongly connected to perceived usefulness, despite students having reservations about reliability (Chung et al., 2026). The preliminary 2026 findings reinforce this – students turn to GenAI partly to make studying more manageable. Among those using AI, 55% said they used it to save time and 41% to reduce stress (AIinHE.org Project, 2026).
Somewhat unsurprisingly, the most common uses were improving writing (69% in 2026) and generating ideas (64%), but 50% of students were also using GenAI to get feedback (AIinHE.org Project, 2026). This matters because feedback plays an important role in student learning. It guides students’ attention to what is important, it shapes their understanding of quality and it directs their future actions.
In other words, half of the students were receiving guidance from LLMs, with all their strengths and weaknesses, about what counts as good quality and how to move forward in their disciplinary studies. Until now, feedback in HE has largely been the domain of educators who bring disciplinary expertise, values and an understanding of not just the current task but also how it supports graduate outcomes. When we explored the reasons, we found that students generally regarded teacher feedback as more helpful and trustworthy, but they also valued GenAI feedback because it was readily available, understandable and less intimidating (Henderson et al., 2026). We should consider these reasons carefully. GenAI was meeting needs that existing university arrangements did not always satisfy.
AI’s influence also extends to how students engage with course readings. While it is perhaps understandable that students seek help in high stakes tasks like preparing for assessments, it is noteworthy that they are also using GenAI to support their reading. In our preliminary 2026 findings, 57% of respondents used GenAI to summarise their course materials, with 31% doing so “often” or “always”. Encouragingly, 71% also reported “often” or “always” reading the original text, while only 8% said they never or rarely did. However, 42% reported reading the GenAI summary first (AIinHE.org Project, 2026). This matters because the summary is likely to influence what they notice, overlook and understand before they encounter the source. In other words, GenAI may have a large influence on how students make sense of their course content, without oversight from their educators.
To some extent, students themselves recognise these tensions in using GenAI in their studies. When we analysed 4,082 students’ open text responses from the 2024 survey, we observed that students described GenAI as helping them understand difficult material, while also expressing concern about its effects on their writing, critical thinking and practical skills. For example, some students reported that they would start with an AI-generated outline and then refine it, but they also recognised that this may reduce their own research efforts and creative thinking. Similar concerns were identified in the 2026 survey, in which students reported limiting their use because of inaccuracies, personal learning values and ethical concerns with 40% of the students expressing concern about their reliance on GenAI (AIinHE.org Project, 2026).
These students’ concerns about their reliance on GenAI echoes a growing body of research suggesting that AI’s immediate benefits may come at a cost to how students engage with learning. For example, Fan et al. (2025) found that with ChatGPT assistance students were able to achieve improved essay scores, but there was no significant improvement in knowledge gain or transfer to a different topic – concluding that GenAI may result in greater dependence and metacognitive laziness. A working paper by Strömberg et al. (2026) tracking more than 26,000 Chinese secondary students, reported faster and improved homework performance but poorer subsequent examination results following GenAI adoption. Studies such as these give us reason to question whether the immediate benefits students experience are always serving their longer-term learning.
The concern extends to the guidance itself, which can be biased and direct students in different directions depending on gender and other identified or inferred characteristics. For example, Du et al. (2027) found that changing gender cues in student essays or prompts could produce systematic differences in feedback: male-coded prompts elicited autonomy-supportive feedback while female-coded prompts elicited more directive and controlling language. What may appear helpful feedback, can misdirect or perpetuate inequity. The risk is perhaps exacerbated by the fact that some students perceive AI’s comments as objective or impartial (Henderson et al., 2026), and therefore may be less likely to be critical or aware of the potential bias.
This brings us back to AI’s influence over attention and future action. Feedback signals which weaknesses matter, which ideas deserve development and what students should attempt next (Henderson, 2019). Similarly, summaries of course materials signal which ideas are important. If students repeatedly use these judgements to guide their own learning, GenAI may influence both what they understand and what they decide to do next. Educators may have little visibility of this guidance or the assumptions behind it.
For educators, perhaps a practical starting point is to recognise that GenAI is already part of many students’ learning, including how they use feedback and approach readings. We need to bring these practices into conversations with students: what are they asking GenAI to do, why are they seeking that help, and how is it shaping their work? An example may be comparing GenAI feedback with an educator’s comments, or an GenAI summary with students’ own readings, could help make disciplinary judgements explicit and prompt students to consider what they are learning and what they may be bypassing. Such conversations could also reveal where university arrangements leave students struggling to find timely, understandable support.
Understanding why and when students turn to GenAI offers insight into how they experience their education and where institutional provision may need to change. We cannot afford to overlook how much of that experience is now mediated by systems that help students interpret course content, judge their work and choose what to do next. As students navigate the knife’s edge between GenAI enablement and dependency, universities cannot afford to leave them to find their own balance.
________________________________________________
Professor Michael Henderson is Professor of Higher Education and Digital Futures, Faculty of Education, Monash University, Dr Jack Walton, CRADLE, Deakin University, Associate Professor Thomas Corbin, School of Education, Deakin University. Michael, Jack and Thomas are members of the AIinHE.org research initiative.
________________________________________________
References
AIinHE.org project: 2026 emerging insights: 10,000+ students’ perspectives of AI and university experience [Conference handout]. (2026). AIinHE.org. https://aiinhe.org/wp-content/uploads/2026/06/2026-hedx-conference-260617_0114.pdf
Chung, J., Henderson, M., Slade, C., Liang, Y., Pepperell, N., Corbin, T., Walton, J., Yu, A. S., Bearman, M., Buckingham Shum, S., Fawns, T., McCluskey, T., McLean, J., Oberg, G., Seligmann, A., Shibani, A., Bakharia, A., Lim, L.-A., & Matthews, K. E. (2026). The use and usefulness of GenAI in higher education: Student experience and perspectives. Computers and Education Open, 10, Article 100347. https://doi.org/10.1016/j.caeo.2026.100347
Du, Y., Borchers, C., & Cukurova, M. (2027). A benchmark for gender bias in large language model feedback on student essays. In E. G. Blanchard, G. Chen, M. Chi, & S. Isotani (Eds.), Artificial intelligence in education (Lecture Notes in Computer Science, Vol. 16586, pp. 163–178). Springer. https://doi.org/10.1007/978-3-032-29773-0_12
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
Henderson, M., Ajjawi, R., Boud, D., Molloy, E. (2019). The impact of feedback in higher education: Improving assessment outcomes for learners. Palgrave Macmillan.
Henderson, M., Bearman, M., Chung, J., Fawns, T., Buckingham Shum, S., Matthews, K. E., & de Mello Heredia, J. (2026). Comparing generative AI and teacher feedback: Student perceptions of usefulness and trustworthiness. Assessment & Evaluation in Higher Education, 51(5), 863–878. https://doi.org/10.1080/02602938.2025.2502582
Strömberg, D., Lei, V., & Wu, Y. (2026). The generative AI learning penalty: Evidence from Chinese secondary education [Preprint]. SSRN. https://doi.org/10.2139/ssrn.6868618