What’s of interest? How to Spot AI Hallucinations Like a Reference Librarian – LLRX
More details (Unless otherwise noted, this description is copied directly from the tagged web site)
Last Tuesday, a client sent me their “thoroughly researched” white paper on workplace automation. It had 47 citations. Looked bulletproof. Every claim backed by a study, every statistic sourced to a journal. I was impressed for exactly three minutes.
Then I tried to find “Peterson et al. (2024): Longitudinal Analysis of Remote Work Productivity in Tech Sectors” in the Journal of Organizational Psychology.
The journal exists. The year 2024 existed. Dr. Peterson probably exists somewhere. But this study? Complete fabrication. ChatGPT had invented a plausible-sounding citation out of thin air. And it wasn’t alone. Of those 47 citations, 31 were what we call “hallucinations.”
The client was mortified. I wasn’t surprised. Because here’s what nobody tells you about AI-generated citations: they’re often more believable than real ones.
This is one among many items I regularly tag in Pinboard as oegconnect, and automatically post tagged as #OEGConnect to Mastodon. Do you know of something else we should share like this? Just reply below and we will check it out.
Yes, citation hallucination is a genuine limitation of generative AI, including tools such as ChatGPT isnt’t it? When these systems are asked to produce references from memory, they can become very confident storytellers, assembling titles, authors, journals, and dates that look entirely plausible.
In my case, I have found Gen AI extremely useful for locating articles for my research, much faster than when I used Google Scholar searches, provided I tell it to search for verifiable sources rather than simply generate a bibliography. I ask GenAI to include the DOI or a direct source link, then open the article and confirm the authors, title, publication details, and whether the study actually supports the claim being made. A DOI is a useful checkpoint, but could be false as well. The DOI needs be checked individually to see if the link goes to the correct article.
Checking citations has always been a standard part of responsible research, with or without AI. So I would frame this “problem” as both an AI limitation and a research-process failure. ChatGPT should never be treated as the final authority, but neither should 47 polished-looking citations make it into a white paper without someone opening them. AI can accelerate the search. It cannot take responsibility for the scholarship.
Also, FYI I am getting a Cloudflare block when I try to access the link in the article on Chrome and Edge. Not sure if it is my system or the host.
Absolutely. I think this is a very important topic for librarians and information professionals, and the example is particularly relevant to our work.
The key point I would emphasize is that AI hallucinations are not simply “wrong answers.” The more dangerous problem is credible-looking misinformation: fabricated articles, authors, journals, DOIs, statistics, and citations that can pass a quick visual inspection.
For librarians, this reinforces a role that is becoming even more important in the AI era: verification, source evaluation, bibliographic literacy, and teaching users how to interrogate AI-generated information rather than simply trust it.
If you want to respond to the OEG post, I would suggest something like this:
Message
This is a very important reminder of the continuing role of librarians in the age of AI. What concerns me most is that AI hallucinations are not always obviously false; fabricated references can look remarkably credible and may even include real authors, journals, and plausible publication details.
This makes information literacy and reference skills more important than ever. Librarians can play a critical role in helping students and researchers verify AI-generated citations, trace claims back to original sources, and distinguish between plausible information and reliable evidence.
Perhaps one of the most valuable skills we can teach in the AI era is not simply how to use AI, but how to question, verify, and critically evaluate what AI produces.