ChatGPT per the case record. The court suspected AI use and issued a show-cause order.
Haneef Muhammad, proceeding pro se, sued Gap Inc. and several other defendants over claims stemming from a shoulder injury, disability benefits, and his termination. The case stalled in the motion-to-dismiss phase as Muhammad filed dozens of motions, twenty-nine of them in under six weeks.
Reviewing those filings, the court found citations to cases that either did not appear to exist or that quoted language the court could not locate in the cited decisions. In his Motion for Recusal, Muhammad cited "Scott v. City of Columbus Legal Department, No. 2:21-cv-5543, 2022 WL 2828015 (S.D. Ohio July 20, 2022)" and a Humphrey case as "controlling authority"; the court found neither exists. He also claimed the court relied on "McDonald's Corp. v. Franklin County" when the case actually cited was In re McDonald, and he attributed a "uncompensable by money" quote to Friendship Materials, Inc. v. Michigan Brick, where that language never appears and the case concerned a Sherman Act claim.
Offered for: controlling authority on the claim
The court granted the defendants' motions to dismiss and dismissed Muhammad's Amended Complaint without prejudice. It held the remaining motions in abeyance and ordered Muhammad to show cause in writing by July 11, 2025, signed under penalty of perjury, why sanctions should not be imposed for his reliance on apparently fabricated cases, and set a hearing for July 16, 2025. The court warned that it was considering sanctions up to and including denial of his motions, dismissal of the action, designation as a vexatious litigator, revocation of his right to proceed pro se, and monetary sanctions.
Kingsfield returned REJECT. A nonexistent S.D. Ohio case cited as controlling. The plaintiff also claimed the court relied on a case that does not exist.
This matter is among the US case-law cases we ran Kingsfield against in the Charlotin benchmark. Kingsfield's existence check confirms every citation against the case-law corpus and an independent source (CourtListener) before any claim resting on it can stand. Across the run Kingsfield ruled 696 of 700 testable claims correctly, including every one of the 504 fabricated citations and 92 fabricated quotes.
Honest scope. This measures Kingsfield's ruling layer given an extracted citation, not an end-to-end run over a full brief. The case record and the AI-tool attribution come from Damien Charlotin's public database and the court order; the verdict is Kingsfield's. See the full 700-claim benchmark.