AI writing tools produce references that look perfect — plausible authors, a real journal, a well-formed DOI — for works that do not exist. The pattern is now documented across courts and scholarship:
Two New York attorneys submitted a brief citing six cases invented by ChatGPT, complete with fabricated quotes and docket numbers. The court sanctioned them and the episode became the canonical warning for AI-assisted legal drafting.
A lawyer for Michael Cohen relied on citations generated by an AI chatbot in a federal motion; the cited cases did not exist. The court called the episode "embarrassing" — the filings, and the coverage, are public record.
A court declaration filed by a prominent misinformation researcher in a Minnesota case was found to contain AI-hallucinated academic citations — references to studies that were never published. The irony made international news; the declaration's credibility did not survive it.
Journals and reviewers increasingly run reference spot-checks, and papers have been retracted or rejected when cited works could not be located. Unlike a legal sanction, a quiet desk reject is invisible — you rarely learn that a phantom reference was the reason.
Every incident above would have been caught by checking each reference against the registries of record before submission. That is the entire product: paste your bibliography, get a verdict on every line.