Practical guide

AI fake case citations: how they arise and how to catch them

Last reviewed November 4, 2026

Quick answerAI tools can cite cases that do not exist, and in Mata v. Avianca the court found bad faith and sanctioned the two lawyers and their firm. Look up every cited case in a case-law source before you file.

AI fake case citations: what they are

An AI fake case citation is a case name and reporter cite that look real but point to no court opinion. In Mata v. Avianca, the lawyers submitted non-existent judicial opinions with fake quotes and citations created by ChatGPT.

Stanford HAI's May 2024 report says its previous study of general-purpose chatbots found they hallucinated between 58% and 82% of the time on legal queries. The study did not measure any current tool.

We may earn a commission if you buy through links on this page, at no extra cost to you.

Screen a brief with CaseRead →

How to catch an invented case before filing

  1. List every case, statute and quotation the draft cites.
  2. Look up each cited case in a case-law source and confirm the opinion exists under that name and cite.
  3. Read the opinion and confirm it says what the sentence claims.
  4. Run a screen if you want a flag list: it marks cites it cannot find and cites that exist but may not back the claim.
  5. Confirm each case is still good law with a citator, KeyCite in Westlaw Advantage or Shepard's in Lexis+ with Protégé.

A "not found" result means the screen's lookup returned no match, not proof that the case was invented. CaseRead's own page says to review those citations by hand before relying on them.

What the Mata v. Avianca and Garner orders found

In Mata v. Avianca, Inc., No. 22-cv-1461 (S.D.N.Y.), the opinion and order on sanctions of June 22, 2023 says there is nothing inherently improper about using a reliable AI tool for assistance. It also says existing rules give attorneys a gatekeeping role over the accuracy of their filings. The court found bad faith on the part of the individual respondents and imposed a penalty jointly and severally on the two lawyers and their firm.

In Garner v. Kadince, 2025 UT App 80 (No. 20250188-CA), the Utah Court of Appeals' per curiam opinion filed May 22, 2025 records that a petition contained fabricated legal authority obtained from ChatGPT and that counsel signed and filed it without independently checking its accuracy. It found a violation of rule 40 of the Utah Rules of Appellate Procedure and ordered fees paid, fees refunded and a donation.

The court said every attorney has an ongoing duty to review the accuracy of court filings. Read your own court's rules and orders.

How many cases have involved invented citations

Damien Charlotin's AI Hallucination Cases Database lists 2,265 cases identified so far, on its page last updated November 4, 2026. By its own description it counts only cases where a court or tribunal explicitly found, or implied, that a party relied on hallucinated content or material. The count moves as cases are added.

Tools that screen a brief for invented cases

CaseRead's Hallucination Shield flags a citation it cannot find and one that exists but may not back the claim; it comes with every plan, Lite ($15 a month, billed monthly) through Team, and its terms leave verifying every citation to the attorney.

Legal citation checker tools compares it with citators and Clearbrief, and the CaseRead free trial guide covers the free check and the 7-day trial on Pro and Team. The four passes are in how to check AI citations before filing, and AI legal research hallucinations covers the study figures.

Sources used for this page

The facts above come from the pages below, read on November 4, 2026. Prices and terms change, so confirm them on the vendor's site before you buy.

  1. Mata v. Avianca, Inc., No. 22-cv-1461 (S.D.N.Y.), Opinion and Order on Sanctions, June 22, 2023 — Independent · storage.courtlistener.com · read November 4, 2026
  2. Garner v. Kadince, 2025 UT App 80, per curiam opinion filed May 22, 2025 — Independent · legacy.utcourts.gov (court address as given on CaseRead's Shield page; text from a public copy of the opinion) · read November 4, 2026
  3. AI Hallucination Cases Database — Independent · damiencharlotin.com · read November 4, 2026
  4. AI on Trial: Legal Models Hallucinate in 1 out of 6 (or More) Benchmarking Queries — Independent · hai.stanford.edu · read November 4, 2026
  5. CaseRead terms — Vendor legal page · caseread.ai · read November 4, 2026
  6. CaseRead Hallucination Shield — Vendor product page · caseread.ai · read November 4, 2026
  7. CaseRead pricing — Vendor pricing page · caseread.ai · read November 4, 2026
  8. KeyCite — Vendor product page · legal.thomsonreuters.com · read November 4, 2026
  9. Lexis+ with Protégé — Vendor product page · lexisnexis.com · read November 4, 2026
  10. Hallucinating Law: Legal Mistakes with Large Language Models are Pervasive — Independent · hai.stanford.edu · read November 4, 2026