Improving student engagement and research persistence with AI

Much of the public conversation about AI in education centers on the risk that students will use it to avoid reading, outsource thinking or produce work detached from sources. Those concerns point to a related question: what changes when AI is designed to operate within scholarly content rather than around it?

A new report from Clarivate, How students are using Chat with the document, examines that question through observed student behavior in ProQuest Research Assistant. More than 470,000 unique users have used the Research Assistant feature on the ProQuest document page, generating nearly 2 million document level interactions. This scale allows for a meaningful examination of how AI affects student engagement with scholarly content. Further, the Chat With the Document feature has been available for more than six months, providing usage patterns and user feedback that reveal how students are interacting with AI.

AI Features Support Thoughtful Reading and Evaluation

The analysis within the report focuses on how students behave when AI is embedded directly in the document view, where reading, evaluation and judgment already take place. Results show that students who encounter Research Assistant features are more likely to engage with documents and act on them.

• • When Research Assistant is visible in the document view, users are 31 percent more likely to scroll through the full text.
  • When the feature is present, users are 40 percent more likely to take meaningful actions such as citing, downloading, saving, emailing or printing a document.
  • Engagement increases further when students interact with AI-generated summaries: Users who clicked to view summaries or key takeaways in full were 76 percent more likely to take meaningful actions with the document.

At a high level, this suggests that the Assistant supports more deliberate decisions about what to read and how closely.

For librarians concerned that AI tools might encourage superficial engagement, the report shows how certain implementations can help students identify which sources warrant their limited time and attention.

Seeing Familiar Research Struggles Reflected Back

The results also show how closely student questions mirror long standing research challenges.

When students use the Chat With the Document feature, their questions tend to cluster around points where research often slows down. Chat is used as a form of support at moments of uncertainty: judging relevance, interpreting complex language, locating evidence in lengthy texts and understanding research design or methodology. Students use it to orient themselves in unfamiliar material, check their understanding, work through dense arguments and keep moving forward. Notably, their questions remain tied to the document at hand rather than drifting toward answer generation.

Responsible Design is Key

A central theme of the report is that responsible use is shaped through design choices, including constraints, transparency and interface decisions.

Document grounded AI confines responses to the material a student is actively reading. Answers are tied to specific passages, reinforcing context and attribution. When students ask the tool to go beyond its intended scope, guardrails step in. Even then, user feedback remains strongly positive, suggesting that clearly explained boundaries can support learning rather than hinder it.

This finding is especially relevant for librarians who are often asked whether AI tools can be “trusted” to uphold academic integrity.

The Help Students Value Most

An interesting takeaway from the report is what students say they value most from Chat With the Document: help evaluating documents, greater clarity, improved relevance and time saved within the research process. They describe the AI tool as a way to stay oriented, reduce confusion and keep working through difficult material. In that sense, the tool appears to support persistence rather than replacement.

For libraries navigating institutional conversations about AI adoption, this reframing shifts the discussion away from efficiency alone and toward learning outcomes librarians already prioritize.

Explore The Report

For academic librarians, the report’s findings provide an evidence based view of how students behave when responsibly developed AI is embedded in trusted scholarly environments. It’s helpful background for conversations with faculty and administrators, while also reinforcing the role libraries play in shaping how new technologies are introduced and evaluated.

Read the full report to explore what document centered AI reveals about how students read, evaluate and persist with research. For more background, learn about our approach to developing responsible AI tools.