Q-Verity catches hallucinated legal citations before they reach the court. One click. 99.9% accuracy. Every citation verified against real case law in real time.
In June 2023, attorneys Steven Schwartz and Peter LoDuca were sanctioned $5,000 by Judge Kevin Castel in Mata v. Avianca for submitting a brief with six fabricated case citations generated by ChatGPT. The cases sounded real. The citations looked real. None of them existed. Since then, the problem has accelerated.
The question is no longer whether you need to verify AI-generated citations. It's how.
Q-Verity is a browser extension that sits inside ChatGPT and verifies every legal citation in real time. No copy-pasting. No switching tabs. No manual Westlaw lookups.
Write your brief, memo, or motion in ChatGPT. Change nothing about your workflow.
One button, right in the ChatGPT interface. Q-Verity scans every citation in the response.
Each citation is cross-referenced against CourtListener's database of real court opinions in seconds.
Each citation flagged as verified, mismatched, not found, or contradicted — with an explanation of why.
| Check | What it catches |
|---|---|
| Citation existence | Fabricated case names and citations that don't exist in any court database |
| Case name match | Cases where the parties don't match the citation — wrong names, swapped plaintiff/defendant |
| Date verification | Cases cited with the wrong year or from the wrong court |
| Holding verification | Claims about what a case held that contradict the actual opinion — the subtlest and most dangerous hallucination |
| Negative treatment | Cases that have been overruled, reversed, or superseded since the date cited |
Any tool can check if a case exists. Only Q-Verity measures the semantic alignment between what the AI claims a case held and what the court actually held — detecting subtle misstatements like "applied strict scrutiny" when the court actually applied rational basis review, or "upheld the statute" when it was struck down.
Q-Verity uses a patented measurement technique that analyzes the geometric relationship between a legal claim and its source document across 144 dimensions of a neural cross-encoder's internal representation.
AUROC under 5-fold cross-validation on 29,887 legal and factual claims across four benchmarks. Near-perfect separation of truthful from hallucinated citations.
Each citation is measured across 144 interpretable dimensions — from surface-level token similarity to deep semantic alignment. Every flagged citation comes with an explanation of where the divergence occurs.
Correct references produce signal. Random references produce none. The detection is not a statistical artifact — it measures real semantic structure.
Built on peer-reviewed research
Developed by the European Institute of Science in Management. Validated on 29,887 items across four benchmarks under 5-fold cross-validation. Protected by international patent filings.
All plans include a 14-day free trial. No credit card required.
Courts are moving from voluntary AI disclosure to mandatory certification. Q-Verity is the verification layer between your attorneys and a sanctions order.
Every citation check is logged with timestamp, result, source, and the specific verification performed. Defensible records for court certification requirements.
Generate the verification record that courts require with one click. Formatted for attachment to filings under standing orders requiring AI disclosure.
See which attorneys are using AI, how many citations are being generated, and whether they're being verified — before filings go out the door.
The sanctions in Mata v. Avianca were $5,000. The reputational damage was incalculable. Don't be the next headline.
Start Your 14-Day Free TrialNo credit card. No installation beyond a Chrome extension. Working in under two minutes.