CiteClear

Fake Legal Citation Checker

Detect made-up legal citations from AI tools. Identify hallucinated cases, statutes, and regulations before they cause problems.

Use Citation Checker How It Works

How AI Hallucinates Legal Citations

🎭 Pattern-Based Generation

AI models generate text based on statistical patterns, not factual retrieval. They can create realistic-looking citations that don't exist.

📖 Mixing Real and Fake

Models often combine real case names with fabricated citations, making hallucinations harder to spot.

📅 Future and Impossible Dates

Citations with future dates or impossible reporter volumes are clear red flags.

🏛️ Non-Existent Courts

References to non-existent courts or jurisdictions often indicate AI fabrication.

Red Flags to Watch For

⚠️ Future-Dated Citations

Cases cited with dates beyond the current year are impossible and clearly fabricated.

⚠️ Impossible Reporter Volumes

Federal Reporter 4th (F.4th) volumes above ~100 are suspicious. F.3d above ~999 doesn't exist.

⚠️ Unrecognized Court Abbreviations

Court abbreviations like "9th Cir." are valid, but "Fed. App." or "Sup. Ct. USA" may be fabricated.

⚠️ Page Numbers Exceeding Reporter Range

Page numbers above ~999 for most reporters are extremely rare and potentially fake.

⚠️ Non-Existent Statute Titles

US Code has 54 titles. References to Title 55+ are fabricated.

⚠️ Inconsistent Formatting

Real citations follow consistent Bluebook/ALWD formats. Inconsistent formatting suggests AI generation.

Common AI Hallucination Patterns

✋ "Case v. Case"

AI often generates placeholder names like "Smith v. Jones" with fabricated citations. Real cases have specific parties.

✋ Sequential Citations

Multiple citations with sequential numbers (e.g., 123 F.3d 100, 123 F.3d 101, 123 F.3d 102) may be fabricated.

✋ All Leading Plaintiffs

If every cited case has the plaintiff's name first alphabetically, it may indicate AI generation bias.

✋ Perfect Citation Balance

Human briefs have uneven citation distribution. AI often creates perfectly balanced citation patterns.

Verification Checklist

1. Check the Court

Verify the court abbreviation exists (e.g., "9th Cir." = valid, "Fed. App. Ct." = suspicious).

2. Check the Reporter

Confirm the reporter abbreviation is valid (F., F.2d, F.3d, F.4th, U.S., S.Ct., L.Ed., etc.).

3. Check the Volume

Verify the volume number is within the valid range for that reporter series.

4. Check the Year

Ensure the year is not in the future and is consistent with the reporter series.

5. Search Online

Use Google Scholar, CourtListener, or official court websites to verify existence.

6. Check the Pin Cite

Verify that the pin citation (page number) is valid for the cited case.

Real-World Examples

✅ Real Citation

Brown v. Board of Education, 347 U.S. 483 (1954)

Verified Supreme Court case. All components are valid.

❌ AI Hallucination

Smith v. Johnson, 999 F.4th 9999 (9th Cir. 2025)

Multiple red flags: F.4th volume 999 doesn't exist, page 9999 is impossible, and 2025 may be in the future.

✅ Real Citation

42 U.S.C. § 1983

Valid statute. Title 42 exists, Section 1983 exists.

❌ AI Hallucination

99 U.S.C. § 9999

Title 99 doesn't exist. US Code only has 54 titles.

Related Tools

Frequently Asked Questions

How do AI models hallucinate legal citations?

Large language models generate text based on patterns learned during training, not by retrieving facts from a database. They can create realistic-looking citations for cases that don't exist by combining real-looking components in ways that seem plausible but are fabricated.

What are the most common types of AI citation hallucinations?

The most common are: (1) completely fabricated cases with real-looking citations, (2) real case names with fabricated citations, (3) citations to non-existent statutes or regulations, (4) citations with future dates, and (5) citations with impossible reporter volumes or page numbers.

Can this tool guarantee a citation is real?

No tool can guarantee a citation is real without checking primary sources. However, this tool uses deterministic pattern matching to flag citations that are definitely invalid (future dates, impossible volumes) or highly suspicious (unrecognized reporters, non-existent courts).

How accurate is the detection?

The tool is highly accurate at detecting obviously fake citations (future dates, impossible volumes). It's less certain about citations that follow valid formats but may still refer to non-existent cases. Always verify suspicious citations through official sources.

What should I do if the tool flags a citation as suspicious?

Treat it as a red flag requiring manual verification. Use the verification checklist above, and search for the citation on Google Scholar, CourtListener, or official court websites. When in doubt, consult a law librarian or attorney.

Can AI models hallucinate citations that pass these checks?

Yes. A sophisticated hallucination could use valid court abbreviations, reporter series, volume numbers, and page ranges while still referring to a non-existent case. That's why manual verification through primary sources is always recommended for critical citations.

Important Disclaimer

This Is Not Legal Advice

This tool provides deterministic format validation and risk triage for educational purposes. It does not:

  • Provide legal advice or legal opinions
  • Guarantee the accuracy of any citation
  • Replace professional legal review
  • Create an attorney-client relationship

Always verify critical citations through official primary sources and consult with a licensed attorney.