Published on

95% of UK Students Use AI β€” But Universities Still Haven't Caught Up

"They told us not to use AI, but not using it felt like falling behind."

That comment from a first-year student in the UK captures something real. And in March 2026, the Higher Education Policy Institute (HEPI) put numbers to that feeling. Their third annual Student Generative AI Survey makes one thing unmistakably clear: students have already decided. The question now is whether universities are ready to respond.

Contents

  1. 95%: AI Is Now the Default in Student Life
  2. A Polarised Landscape β€” Neither Encouraging Nor Banning
  3. Arts and Humanities Students Are Being Left Behind
  4. Does AI Make Students More Lonely or Less?
  5. What This Data Is Really Asking

1. 95%: AI Is Now the Default in Student Life

HEPI has tracked AI use among UK undergraduates each year since 2024. The trajectory is unmistakable:

  • 2024: 66%
  • 2025: 92%
  • 2026: 95%

A 30-percentage-point rise in two years. The survey β€” the third in the HEPI/Kortext series β€” was based on 1,054 full-time undergraduates in the UK. "Use AI" here does not mean trying it once; it means integrating AI into academic work in at least one way.

In terms of tools, ChatGPT ranked first by a significant margin, followed by Microsoft Copilot, Google Gemini, Perplexity, Grammarly, and Turnitin.

On institutional provision: the share of students whose institutions provide generative AI tools rose from 9% in 2024 to 23% in 2025 to 38% in 2026. That growth is real β€” but it also means six in ten students are still sourcing and using AI tools entirely on their own.

2. A Polarised Landscape β€” Neither Encouraging Nor Banning

The word HEPI chose to describe 2026's picture is "polarised." They used it explicitly in their report, and the data backs it up.

On whether their institution encourages AI use:

  • Students who feel encouraged: 37%
  • Students who feel not encouraged: 36%

Almost exactly half and half. The same campus, the same year β€” and students are receiving entirely different signals depending on which course or lecturer they encounter. Without clear institutional direction, students are left to form their own standards.

The most notable shift was at Russell Group institutions (the UK's 24 leading research universities). In 2025, Russell Group students were among the least likely to feel their institution encouraged AI. By 2026, they topped the list at 39% β€” a 13-percentage-point increase in a single year. The more research-intensive universities appear to be shifting first. But that shift has not yet spread evenly across the sector.

3. Arts and Humanities Students Are Being Left Behind

68% of students believe AI skills are essential to thrive in today's world. But only 48% feel their teaching staff are actively helping them develop those skills.

That gap is widest in Arts and Humanities. These students are the most likely to feel under-supported by their educators when it comes to AI. This is striking because AI has penetrated deeply into the very tasks these students do most β€” writing, textual analysis, critical argument construction. Yet precisely in these fields, AI literacy education is lagging most.

The underlying assumption β€” that AI is mainly relevant to STEM students β€” is creating a structural blind spot. Students in humanities and creative disciplines use AI too, often to do what their degree is supposed to teach them. Without guidance, they have no framework for using it well or for recognising its limitations.

4. Does AI Make Students More Lonely or Less?

It sounds like an unexpected question, but it generated one of the survey's most striking findings.

  • AI makes me feel more lonely: 20%
  • AI makes me feel less lonely: 21%

Almost perfectly split. For some students, AI is a resource that gives them answers when nothing else does β€” a kind of availability that human teachers cannot always provide. For others, it is a screen-bound interaction that replaces rather than supplements human connection.

The split extends to how students source information:

  • Prefer traditional sources (books, academic papers): 33%
  • Use equal balance of AI and traditional sources: 29%
  • Prefer AI sources: 37%

AI preference has overtaken traditional-source preference. But without instruction in how to verify and critically evaluate AI outputs, students are making that shift without the tools to do it well.

5. What This Data Is Really Asking

Three years of HEPI surveys tell a consistent story: students have moved on. AI is part of how they study, think, and prepare for working life. The question that remains is what universities are doing in response to the choices students have already made.

The 37%-versus-36% split in institutional attitudes is not neutrality. It is incoherence β€” a policy vacuum that produces inconsistency within and across institutions. The consequences are real: an under-served cohort in the humanities, confusion about what is and is not acceptable, and students left to work out the rules of AI use by themselves.

95% is a starting point. The more important question is: what are universities teaching that 95%?

Sources

Discuss on Twitter β€’ View on GitHub
95% of UK Students Use AI β€” But Universities Still Haven't Caught Up | MINSSAM.COM