How Much Are Faculty Using AI for Scholarly Discovery?

p strongFaculty use of AI for research is real but fragmented: broad tools attract regular attention, while specialized scholarly-discovery services still reach relatively small shares of the academic workforce./strong/p p In Primary Research Group's survey of 660 US higher-education faculty, respondents reported a mean of 37 minutes per week using ChatGPT. By contrast, only 4.55% had ever used Semantic Scholar. The gap shows why libraries cannot assume that familiarity with generative AI automatically translates into familiarity with purpose-built research tools./p p Adoption also varies by discipline. Among fine-arts and architecture faculty, 30.4% reported routinely using AI to search for journal articles or other content. Faculty in law and in education or library science reported some of the strongest productivity gains. Nearly 26% of male respondents rated AI training a high or very high priority, illustrating that demand for support is present but not uniform./p p The opportunity for academic libraries is to connect tool enthusiasm with research quality. Short demonstrations can compare a conversational answer with a database search, show how to trace a generated summary back to the original article and explain when disciplinary indexes outperform a general model. Research offices can add guidance on confidential proposals, unpublished findings and disclosure expectations./p p The most useful benchmark is not simply minutes spent with AI. Institutions need to know which disciplines use which tools, for what stage of research and with what confidence in the results. That is the information required to design training, licenses and safeguards that faculty will actually use./p p  /p