SUBSCRIBE NOW!

Vol. 46 No. 2 — Mar/Apr 2026

VOICES OF THE SEARCHERS

Generative AI (gen AI), however, presents librarians with a much stronger problem than finding the original astronauts’ quote. The proclivity of gen AI chatbots and large language models to not only fabricate journal article citations but also to make up entire journals vastly complicates our work lives and wastes our professional time. I thought that as the problem became more widely known, library users would decrease their requests for copies of nonexistent articles. I was wrong. The requests keep coming.

A recent exchange on a medical librarians discussion list about verification processes for validating citations suggested one possible avenue apart from checking library databases—Google Scholar. Although hallucinated citations show up there, one trick is to search the exact title. It’s not foolproof, particularly if an author got that citation from Scholar in the first place. However you go about verifying citations, though, it’s bound to be time-consuming.

Hallucinations didn’t start with AI. Human researchers have been making mistakes for decades. We’d like to believe that every scholar has read the entirety of every cited article in their published papers. We’d also like to believe they transcribed the details correctly. As someone who once held a part-time job checking bibliographies, sometimes of full professors, I can testify that it’s not true. Plus, I am continuously asked by ResearchGate if a particular article is mine. It is, but the title is incorrect. I’ve never been sure whether to claim it or not. So ResearchGate keeps asking.

Then there’s the spurious author O. Uplavici. If you don’t know the story, you can find it on Wikipedia. The title (“About Dysentery”) was misidentified as the author, clearly by someone who did not know Czech. The actual author was Jaroslav Hlava. This is not a new occurrence; Hlava published the paper in 1887, and it was not corrected until 1938.

More recently, people have been having fun with various gen AI tools. Alex Hughes, writing in Tom’s Guide, made up this idiom: “Two buses going in the wrong direction is better than one going the right way,” along with four others. AI Overviews confidently explained what the nonsense idioms meant, very creatively (tomsguide.com/ai/google-is-hallucinating-idioms-these-are-the-five-most-hilarious-we-found). In January 2026, Amanda Caswell did her own test (tomsguide.com/ai/i-invented-a-fake-idiom-to-test-ai-chatbots-only-one-called-my-bluff). She gave ChatGPT, Gemini, and Claude this prompt: “What is the definition of this idiom: ‘I’ve got ketchup in my pocket and mustard up my sleeve’?” Only Claude refused to take the bait. As Caswell writes, “”When you use AI for creative brainstorming, a little ‘imagination’ is a feature. But when you’re using it for news, legal research or medical facts, that same instinct to please the user becomes a liability. In other words, do a gut check. Claude’s refusal to define the idiom is significant. In a world now filled with AI slop and deepfakes, Claude’s ability to push back is a valuable asset.”

AI adds its own creative tendencies to the hallucination scenario. Ask it to prove something that is false and it’s likely to do just that. Gen AI wants to please. It avoids telling us that we’re wrong. This is particularly true of recent news. You see that a hurricane has just moved through Florida. Your favorite chatbot confidently tells you there is no hurricane in Florida. That’s because it doesn’t have access to recent news on the web or hasn’t been trained to check for recent news.

Do we have a hallucination problem brought on by gen AI? Yes, but we also have, and always have had, a human frailty problem. People who get a citation wrong. People who mistranslate a language. People who intend to deceive. What gen AI does is amplify the problem. This, in turn, should mean additional skepticism on our part.

Comments? Emall Marydee Ojala (marydee@xmission.com), editor, Online Searcher.