Students don't want AI to think for them

For the past two years, much of the conversation about artificial intelligence (AI) in higher education has centered on how students are using it. Are they using these tools to learn more effectively, or to bypass the work of learning?

Findings from a recent Clarivate survey of more than 500 undergraduate and graduate students from around the world and across disciplines suggest the answer is more nuanced and encouraging than the debate often allows. Ninety-five percent of students reported using AI for school projects or assignments, up from 45% in a similar survey conducted in 2024. Institutional policies, meanwhile, have changed relatively little. The result is a growing disconnect between students’ use of AI and institutions’ efforts to guide it.

Those findings are gathered in Students Are Sprinting Ahead: What 500 Students Told Us About AI in Academic Work, a new report on how students use AI, where they struggle and what they want next.

AI Has Become Part Of The Academic Workflow

Students are using AI across the research and learning process. In the Clarivate survey, 61% reported using AI to summarize articles and 61% for brainstorming research ideas. More than half use it to prepare for exams, while 51% use it for writing and editing papers. Nearly half use it to find scholarly sources.

The way students begin research is changing as well. Thirty-one percent start a project in a general AI assistant such as ChatGPT, Claude or Gemini. Only 10% begin at the library website or database. Another 9% start in a dedicated AI research tool, a category that barely existed a few years ago.

For academic libraries, this represents an important shift. Increasingly, students are entering the research process through conversation rather than search.

Students Aren't Blindly Trusting AI

With concerns about misinformation and hallucinations linked to AI, students appear more thoughtful about AI's limitations than many assume.

About 60% students reported fact-checking AI-generated content “most of the time” or “always.” That percentage jumps to 64% for graduate students.

Students also reported on what helps them trust AI-generated academic content: source links, citations and references. In fact, original source links and bibliographies ranked as the most important trust-building features in an AI tool.

The Real Issue Isn't Speed. It's Trust.

Students also identified areas where AI tools were less effective.

Some reported inaccurate or fabricated citations. Others said they spent significant time verifying AI-generated information. In these cases, AI did not eliminate the work so much as shift it toward checking and correction.

For that reason, trust emerged as a central theme in the research.

When asked what they wanted from an ideal AI tool, students most often pointed to trusted, verifiable academic support. They emphasized citations linked to credible sources, transparent evidence and confidence that the information they use can be checked.

Students Want Support, Not Shortcuts

Perhaps the most surprising finding was what students did not ask for.

• • They weren't looking for a tool that could complete assignments without effort. Instead, they described wanting AI that acts more like a mentor than a shortcut.
  • Students said they wanted tools that adapt to their level of understanding, explain concepts in different ways, challenge their reasoning and support long-term learning. Personalized tutoring and guidance emerged as one of the strongest themes in the survey.
  • They also wanted help navigating the research process itself, from discovering literature and identifying gaps to organizing sources and managing citations.

Taken together, the responses point to students’ interest in AI that helps them think better, not simply finish work faster.

What This Means For Academic Libraries

Academic libraries have long helped students evaluate information, verify sources and build research skills. Those responsibilities remain relevant as AI becomes more common in academic work.

The survey suggests that students continue to value many of the same things libraries support: credible information, transparent sourcing, critical evaluation and guidance through the research process.

For libraries, AI presents an opportunity to bring established expertise in information literacy, evidence and research support into the way AI tools are evaluated and used.

To explore the full findings and implications for the future of academic research, read our report, Students Are Sprinting Ahead: What 500 Students Told Us About AI in Academic Work.