An Oklahoma Judge Let ChatGPT Write His Research, and the Cases It Cited Don't Exist
What happened
When Oklahoma District Judge Lawrence Wheeler signed a court order in September 2026, it contained at least two case citations that cannot be found in any legal database. Wheeler later admitted he had used ChatGPT to conduct the legal research behind the order. The citations, which appeared with the authority of a judicial ruling, referred to cases that had never been decided.
Legal research has always depended on accurate citation. A court order that points to non-existent precedent does not just embarrass the judge who signed it; it corrupts the record that attorneys and parties before the court rely on to understand the law as applied to them. If either party to the proceeding shaped their arguments around those citations, or if the order was used as authority in a subsequent filing, the damage extends beyond a clerical error and into the substance of the proceeding itself.
This is not the first time fabricated case citations have appeared in legal documents attributed to AI output. Several attorneys in prior years faced sanctions after chatbot-generated briefs cited phantom cases before federal courts. Those incidents involved lawyers who, under professional rules, bear explicit responsibility for verifying every citation they submit. A judge occupies a different position. No external adversary checks a judge's research before it reaches the page; the verification stops at the judge's own desk.
Wheeler's admission prompted investigations into the matter and raised the prospect of disciplinary action. Whether that process produces formal sanctions or stops at a reprimand, the underlying question it surfaces is procedural: courts have no standing policy on whether and how judges may use AI tools in the course of research, and no mechanism to flag when they do.
That absence is the structural gap. At present, there is no way to reconstruct what a judge asked an AI system, what it returned, and which parts of the response were verified before going into an order. A provable record of what a system produced, and what steps were taken to check it, would make the difference between a finding of negligence and a finding of nothing at all. Without that record, an order containing a hallucinated citation looks identical to one containing a real one, at least until someone tries to look the case up.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.