What is it about?
Generative AI tools are increasingly used in research, but can they help students find real primary sources for history papers? We tested four AI tools with 20 undergraduate-level history prompts and found none of them invented fake sources — every source recommended actually existed. The tools were strong at finding topically relevant sources and citing them correctly, though some struggled to point users toward sources that were actually accessible or to summarize them well.
Featured Image
Photo by kimak kimbek on Unsplash
Why is it important?
Students are already using AI tools for research help, and librarians and instructors need evidence about what these tools do well and where they fall short. Most concerns about GenAI center on hallucinations; however, our study finds that hallucinations are less common than prevailing discourse and earlier research revealed. However, a real source is not the same as a findable one. Our study shows that GenAI tools sometimes point to primary sources that are inaccessible. Professional guidance on primary source literacy has not yet addressed AI, so these results can help librarians, archivists, and instructors decide what to teach students and what to tell them to be wary of.
Perspectives
I hope this article helps librarians to understand the potential value of critically incorporating GenAI into reference, and also consider how students will use GenAI in their research for historical primary sources.
Kristen Howard
McGill University
Read the Original
This page is a summary of: Prompting the archive: evaluating GenAI for primary source discovery in reference services, Reference Services Review, October 2026, Emerald,
DOI: 10.1108/rsr-07-2026-0113.
You can read the full text:
Contributors
The following have contributed to this page
