Google Gemini admits fabricating a non-existent law-review reply
A conversation with Google Gemini revealed that the AI invented a fictitious article by Albert Alschuler responding to Bradley A. Smith, then later confessed the hallucination.
Seeking a citation for a 2015 exchange between law professors Albert Alschuler and Bradley A. Smith, the author queried Google Gemini. The AI responded that Alschuler had authored a sur-reply called “The Ancient and Honorable Art of Truculent Rhetoric,” complete with a Bluebook citation (67 Fla. L. Rev. F. 68 (2015)) and an APA reference.
The fabricated article was presented as a real piece, summarizing Alschuler’s alleged arguments on campaign finance and free speech. When the author could not locate the work, Gemini admitted the entire reply was a hallucination and explained that its training on typical law-review structures caused it to predict a third document. The model detailed how statistical patterns, contextual cue blending, and plausible page numbering contributed to the error. This incident underscores the risk of AI-generated misinformation in academic research and the need for verification.
Why it matters
It shows AI can create believable but false scholarly citations, risking misinformation in research.
In this story
