A third of all scholarly information discovery is now mediated by artificial intelligence, and many researchers have little awareness of how deeply AI shapes the scientific literature they see. The finding comes from the first phase of Taming the Crocodile, a major industry study by Kudos that surveyed 11,500 researchers and 200 librarians and scanned more than 300 sources.
The Phase I report, delivered to sponsors in July 2026, provides analysis, recommendations, and a strategic roadmap for publishers, librarians, and technology providers. It documents a rapid shift in how researchers begin their searches: one in 10 now start directly in AI tools, while another 25% start with general search engines where AI-generated answers are increasingly embedded. Among early-career researchers and those in corporate settings, the proportion beginning with AI tools doubles.
Misunderstanding and hallucinated references
The study reveals widespread confusion about AI's role. Forty percent of researchers believe that AI-generated overviews in search engines are created or curated by humans, a figure that rises to 46% among students. Librarians reported that users increasingly request articles that do not exist, pointed to hallucinated references generated by AI systems.
Traditional engagement metrics are also feeling the pressure. Librarians observed declines in usage of primary sources as discovery habits change, and many worry that subscription purchasing decisions still lean heavily on those same usage numbers.
What librarians and publishers want
Librarians want publishers to take a more active role in ensuring scholarly content is represented accurately inside AI systems. The report found strong support for greater transparency around publishers' AI engagements and for thoughtful approaches to licensing content for AI training and agentic querying.
One clear recommendation calls for publishers to create structured, plain-language content designed specifically for AI discovery. AI optimisation specialists consistently advise developing "answer-ready" content that explains the importance and relevance of research in clear language. Phase II of the project will focus on industry best practices for schema, metadata, and mark-up to improve AI visibility and attribution.
Industry voices respond
"The findings give us all such a strong foundation to help adapt our industry infrastructure," said Emilie Delquié, Chief Product & Customer Success Officer at Silverchair, which sponsored the study.
Shane Rydquist, Associate Vice President, Delivery and Solutions at CACTUS Communications, pointed to a central tension. "This is the first evidence base we have seen that connects how researchers actually search to what authors and publishers should do about it. The finding that matters most is that researchers rely on AI but do not trust it, and that evidence and clear attribution are what earn both the click and the citation."
Dan Penny, Head of Market Intelligence at Springer Nature, added: "The fast-moving and unpredictable impact of AI on all aspects of scholarly communication makes it essential to stay ahead. We all need to act quickly, and responsibly, on the findings here to ensure that we continue to support researchers, the wider community, and the integrity of scholarly communication."
Why this matters for researchers, educators, and IT professionals
For researchers and librarians, the study underscores a pressing need to understand how AI filters and presents scholarly work. AI for Science & Research is no longer a niche topic-it directly affects which studies get seen, cited, and trusted. Misattributed or hallucinated references can erode the credibility of literature reviews and grant applications, making digital literacy around AI tools a core skill.
For IT and development teams supporting academic institutions or publishing platforms, the push toward structured, machine-readable content and new metadata standards signals a wave of technical work ahead. Publishers who invest now in "answer-ready" formats and AI-friendly markup will be better positioned as discovery shifts further away from traditional search. The report also recommends benchmarking AI visibility and developing educational materials-both areas where cross-functional teams in IT, research support, and education can collaborate.
Your membership also unlocks:
