As students across the country celebrate their A-level results and prepare for freshers' week, I'm reflecting on how Amytis can improve the future of AI in research.

The Higher Education Policy Institute (HEPI) published their Student Generative AI Survey 2026 back in March. The findings paint an interesting picture of how undergraduates are (or in some cases, are not) using AI in their studies. 

As someone who finished university before open source AI launched, some results feel exciting. Others make me question whether students are truly being set up for success.  

Students are (almost) universally using AI, but the best AI tools for research are underused

95% of students now report using at least one form of generative AI in their university work, up from 66% just two years ago. This is a remarkable shift in a very short period.

But what strikes me most is the data on what tools people actually use. More than half of students use AI to generate text, using tools like ChatGPT or Claude. Next is summarising lectures or taking notes (37%) and editing written content (38%). While incredibly useful, these tools are not always the best AI for research purposes.

Notably, the least common use cases are those that do support specific research-based objectives. Only 17% of students use AI for data analysis or to write code. Life science students who only use AI to find academic papers or to do literature reviews are missing out on AI's full benefits.

Amytis fills this gap. It gives students a tool to create and run experimental workflows from scratch using specialised life science models. Students can structure and organise their workflows intuitively without being forced to work in a linear chat.

AI consequences are top of mind, from misconduct to skill erosion

The survey also captures something that often gets lost when talking about AI: students feel anxious about how they use these tools.

The most common deterrents to AI use are fear of being accused of cheating (42%), and getting false (35%) or biased results (32%).

For life science researchers specifically, these stakes are higher than most. A misidentified gene, a mislabelled dataset, or a statistical result that goes unchecked risks both academic and scientific integrity.

There's also a growing concern about skill erosion. Several students in the survey's qualitative responses put it bluntly: "My grades have dropped because AI is mitigating my ability to think critically." Or my favourite quote, "I'm not using my brain at all."

For PhD students and researchers, “not using my brain” is not an option. Intellectual ownership at this level is crucial, and if you can’t explain what your analysis pipeline did, or why it made the choices it made, you're not really in control of your research. 

For the team at Amytis, this is why we want to keep the researcher in control of their workflows. Users can clearly see and edit which prompt each model runs and in what order. We are also developing specialised agents that challenge researchers on their approach. These agents can explain behind-the-scenes processing steps and help users critically interrogate their work.

The AI literacy gap is wider than most universities are admitting

Perhaps the most striking finding is the gap between what students believe they need and what they are actually getting.

68% of students say that understanding and using AI effectively is essential to thrive in today's world. However, only 48% feel that their lecturers are helping them develop those skills for their future careers, and 33% arrive at university having had no prior experience using AI at school.

The students who arrive without prior experience, who receive limited structured guidance once they get there, and who default to general-purpose chat tools for everything are not set up to use AI for research well.

The Amytis AI literacy series

Starting this autumn, we will run a series of workshops and webinars specifically for life science students at university. These sessions will offer practical support on using generative AI, including when and how to use it well. We want to help students amplify their thinking, not replace it.

We will also be providing demos and exclusive joining offers for students who want to be the first to try Amytis in their research.

If you are a student, society lead, or university staff, let's talk! Sign up for our waiting list at amytis.io, follow us on LinkedIn or keep an eye on our Luma calendar to hear about our events first.

Sources:
Rose Stephenson and Charlotte Armstrong, Student Generative AI Survey 2026, HEPI Report 199, March 2026. hepi.ac.uk
Photo by Meredith Spencer on Unsplash