How Judges Are Responding to AI-Hallucinated Citations in Court Filings

Written by the Citation Safe Research Desk · Reviewed by Andy Gaber, Founder — July 16, 2026

Judicial responses to AI-hallucinated citations follow a fairly consistent pattern across the documented cases in the Charlotin AI Hallucination Cases database, even though the specific consequences vary by jurisdiction and severity.

How judges typically discover the problem

Most commonly, opposing counsel raises the issue after being unable to locate a cited case. Less commonly, a judge's own law clerk or research attorney flags it while preparing for a hearing. In a smaller but growing number of documented cases, judges themselves have run a quick database search on an unfamiliar-looking citation and found nothing.

The order to show cause

Once a suspected fabrication is identified, the typical judicial response is an order to show cause, requiring the filing attorney to explain, often under oath or by declaration, how the citation was generated and why it should not result in sanctions. This was the procedural path in Mata v. Avianca, No. 22-cv-1461 (S.D.N.Y. 2023), where Judge P. Kevin Castel required the attorneys to produce the underlying documents and explain the ChatGPT-generated research process in detail.

Escalating judicial patience over time

Judicial commentary in more recent documented cases — including Wadsworth v. Walmart Inc. (D. Wyo. 2025) and Coomer v. Lindell (D. Colo.) — reflects less tolerance for "I didn't realize AI could fabricate citations" as an excuse than earlier cases immediately following Mata. Judges increasingly note in their orders that the risk of AI hallucination is now widely publicized, reducing the persuasiveness of a surprise defense.

Standing orders

A growing number of judges have adopted standing orders requiring attorneys to disclose AI tool use and certify independent verification of any AI-assisted research before filing. See our 2026 overview of judges' standing orders on AI use for a jurisdiction-by-jurisdiction survey.

Expert perspective

Judicial ethics commentators have observed that courts are converging on a consistent expectation regardless of specific local rules: an independent verification step before filing, documented if possible, is becoming the de facto standard against which "reasonable inquiry" under Rule 11 is measured.

A common question

Do judges treat self-reported errors differently from errors caught by opposing counsel?

Yes, consistently. Documented cases show more lenient outcomes for attorneys who identify and correct their own error promptly, compared to attorneys who defend a citation after opposing counsel flags it as fabricated.

A deeper look: how discovery of a fabrication actually unfolds

In practice, judges rarely stumble onto a fabricated citation cold. The far more common path, per the fact patterns in the Charlotin database, runs through opposing counsel's own cite-checking process: a paralegal or associate tasked with responding to a motion tries to pull the cited case for a responsive brief, cannot locate it in Westlaw, Lexis, or CourtListener, and escalates to the attorney of record. That attorney then raises it with the court, typically in a footnote or a separate letter, rather than silently ignoring it. Judges have repeatedly noted in written orders that they rely on this adversarial check functioning correctly, which is itself a reason courts respond severely when it turns out no one, on either side, caught the problem before oral argument or a ruling.

The role of law clerks and staff attorneys

A second, less visible detection path runs through chambers itself. Law clerks and staff attorneys preparing a bench memo or draft order routinely pull every case cited in the briefing to confirm the court's own characterization of the holding is accurate. Several documented cases show the fabrication being caught at this stage, sometimes after briefing is complete and before a scheduled hearing, which is why some show-cause orders are issued with unusually short response windows: the court has already done the confirming legwork and is not asking a hypothetical question.

What this means for your own filings

Assume that at least one of three readers — opposing counsel, a law clerk, or the judge personally — will attempt to pull every citation in your brief. That assumption alone should be enough to make independent verification before filing feel less like an extra step and more like an inevitability you are choosing to get ahead of rather than react to.

A note on appellate review

Judicial responses at the appellate level carry an added dimension: a fabricated citation in an opening brief can taint the entire appeal if the panel relies on it before catching the problem, since appellate courts generally have less opportunity than trial courts to hold live hearings where errors surface through argument. Appellate panels have responded by requesting supplemental briefing, striking the offending sections, or in more serious cases, referring the matter to the circuit's own attorney discipline apparatus. The compressed page limits and tight briefing schedules typical of appellate practice make this an area where AI-assisted drafting is common and verification discipline needs to be correspondingly strict.

A practical takeaway for practitioners at every level

Whichever court you are in, the safest assumption is that a fabricated citation will eventually be checked by someone with both the time and the incentive to find it. Building verification into your workflow before filing, rather than hoping it survives review, remains the only approach the public record supports as reliable.

Final takeaway

Judges are not adversaries in this process, but they are also not going to independently verify your citations for you before ruling against you for relying on ones that don't hold up. Treat the possibility of judicial scrutiny as a certainty rather than a risk, and verify accordingly, every time, regardless of how routine the filing seems.

It is also worth noting that judges themselves are not uniformly tech-savvy about how AI hallucination actually works, and written orders sometimes reflect a still-developing judicial understanding of the underlying technology. This is gradually changing as more judicial education programs address AI directly, but it means the specific reasoning in an order can vary even where the outcome is consistent.