Artificial intelligence tools can draft emails, summarise documents, and answer research questions in seconds. But they can also produce something far more troubling: references to court cases, articles, and studies that simply do not exist. These invented sources are known as AI hallucinated citations, and they have caused embarrassment, sanctions, and professional discipline for lawyers, students, and researchers around the world.
This article on AI hallucinated citations explained covers what they are, why AI tools create them, high-profile court cases, how courts and professional bodies are responding, and practical steps to avoid relying on fake sources.
Table of Contents
What Are AI Hallucinated Citations?
An AI “hallucination” happens when a generative AI tool produces information that sounds accurate but is false or fabricated. A hallucinated citation is a specific type of error in which the AI invents or distorts a source, such as:
- A court case with a realistic name and docket number that was never decided
- A real case paired with a made-up quotation or holding
- An academic article attributed to real researchers who never wrote it
- A book, report, or statistic with a convincing but fictional reference
- A real source cited for a point it does not actually support
Because these citations often follow the correct format and look authoritative, they can be difficult to spot without checking the original source.
Why Do AI Tools Hallucinate Citations?
Most popular AI chatbots are built on large language models. These models are trained on vast amounts of text and learn to predict which words are likely to come next. They are designed to produce fluent, plausible language, not to look up facts in a verified database.
Key Reasons Hallucinations Happen
- Pattern prediction: the model knows what a legal citation or research reference usually looks like and can generate one that fits the pattern
- Gaps in knowledge: when the model lacks real information, it may fill the gap with invented details instead of admitting uncertainty
- Pressure to answer: prompts asking for “cases that support this argument” can lead the AI to produce answers even when none exist
- Mixing sources: real names, courts, and dates can be combined into fictional references
- Outdated or limited data: models may not have access to current or complete legal and academic databases
Even AI research tools connected to legal databases are not immune. A 2025 study cited by legal librarians found that some retrieval-based legal research tools still produced hallucinations at rates of roughly 17 to 33 percent in testing.
The Case That Made Headlines: Mata v. Avianca
The issue gained worldwide attention in 2023 with Mata v. Avianca, a personal injury lawsuit against an airline in federal court in New York. Lawyers for the plaintiff filed a brief citing several court decisions that turned out not to exist. The cases had been generated by ChatGPT.
When the opposing side and the court could not locate the cases, the lawyers initially stood by them. On June 22, 2023, Judge P. Kevin Castel issued a sanctions opinion, finding that the lawyers had acted in bad faith and imposing penalties. The judge even attached one of the fictional opinions to his ruling, illustrating how convincing the fabricated text appeared.
How Big Is the Problem?
Since Mata, the number of reported incidents has grown rapidly. Researcher Damien Charlotin maintains a widely cited database tracking court decisions involving AI hallucinations in legal filings.
| Year | U.S. Instances Recorded | Notes |
|---|---|---|
| 2023 | 11 | Includes Mata v. Avianca |
| 2024 | 37 | Growing awareness among courts |
| 2025 | 524 | 97 monetary sanctions reported |
| 2026 (to late July) | 668 | Already above the 2025 total |
These figures, reported by the Maryland State Bar Association in July 2026, include fabricated cases as well as other AI-related errors such as false quotations and misstated holdings. Different trackers use different methods, so totals vary between sources.
Who Is Submitting Fake Citations?
According to the same data, self-represented litigants account for the largest share of incidents, followed by lawyers. A small number of cases have involved expert witnesses and even judges. Most monetary sanctions have fallen in the range of about $1,000 to $5,000, though a few have been much larger, and disciplinary referrals to bar authorities are increasingly common.
How Courts and Regulators Are Responding
- Sanctions: fines, orders to pay opposing costs, and dismissal of filings
- Disciplinary referrals: reporting lawyers to state bar authorities
- Standing orders: many judges now require parties to disclose AI use or certify that citations were checked
- Court rule changes: in 2026, the Florida Supreme Court amended its rules to require that filings verify that cited legal authorities exist and are accurately cited
- Ethics guidance: the American Bar Association’s Formal Opinion 512 (2024) and state bar opinions remind lawyers of their duties of competence, candour to the court, and supervision when using AI
The core message from courts is consistent: AI does not excuse a professional from verifying every source they submit.
Beyond the Courtroom
Hallucinated citations are not only a legal issue. They have appeared in:
- Academic papers and student essays: fake references can lead to failing grades or misconduct findings
- Journalism: invented quotes or sources can damage credibility
- Government and corporate reports: fabricated references have led to public corrections
- Medical and scientific writing: where inaccurate sources can have serious consequences
Many universities and publishers now ask authors to disclose AI use and to confirm that every reference has been checked against the original source. Editors and teachers are also learning to look for the warning signs of invented references.
How to Spot a Hallucinated Citation
- The case, article, or book cannot be found in official databases or library catalogues
- The docket number, volume, or page does not match the court or journal
- The quoted language does not appear in the actual source
- The author or court exists, but has no record of the cited work
- The citation seems unusually perfect for the argument being made
Best Practices for Using AI Safely
- Verify every citation: check each source in an official database, court record, or library
- Read the source: confirm that it actually says what the AI claims
- Use AI as a starting point: treat outputs as drafts, not final research
- Choose reliable tools: prefer tools connected to verified databases, while still checking results
- Follow court rules: comply with disclosure and certification requirements
- Train staff: law firms, schools, and businesses should teach responsible AI use
- Document your process: keep records of how research was verified
Final Thoughts
With AI hallucinated citations explained, the lesson is clear: generative AI can be a helpful assistant, but it is not a reliable source of truth on its own. From Mata v. Avianca to hundreds of cases since, courts have shown little patience for fabricated authorities. Whether you are a lawyer, student, researcher, or writer, the responsibility to check every citation remains yours.
This article is for general information only and is not legal advice.
