Case Citation Checker: How to Verify a Legal Citation Is Real Before You File

The case count comes from a public database maintained by Damien Charlotin, which logs only matters where a court explicitly found or implied that someone relied on hallucinated material. That makes it a floor, not a ceiling. The trend inside it is the part worth pausing on: more decisions were logged in the first seven months of 2026 than in the whole of 2025.

The hallucination rates come from the peer-reviewed study by Magesh and colleagues at Stanford, which tested the LexisNexis and Thomson Reuters AI research products and concluded that vendor claims of hallucination-free citations were overstated. Those tests ran in 2024 and both products have been revised since, with no independently replicated update published. The durable lesson is not about any one vendor. It is that a legal-grade price tag does not transfer the verification duty to the software.

One shortcut worth knowing, recommended by law librarians: tools that extract every citation from a document and run them through a citator will only return authorities they can find, so an entry that comes back empty is a fast signal that the case may not exist at all. That turns a slow manual pass into a triage step.

The most common misconception is that courts need a new AI rule before any of this is sanctionable. They do not. Federal Rule of Civil Procedure 11(b)(2) says that by signing, filing, or later advocating a paper, you certify that the legal contentions are warranted by existing law, formed after an inquiry reasonable under the circumstances. A case that does not exist cannot survive a reasonable inquiry, which is why every sanction on this page rests on rules that predate generative AI by decades. Courts have mostly acted on their own initiative under Rule 11(c)(3), where the 21-day safe harbor does not apply, alongside 28 U.S.C. section 1927 and their inherent power.

Some judges went further and issued standing orders. Judge Brantley Starr in the Northern District of Texas published the first in May 2023, requiring every attorney appearing before him to certify either that no part of a filing was drafted by generative AI, or that any AI-drafted language was checked for accuracy by a human using print reporters or traditional databases. Judge Stephen Vaden at the Court of International Trade requires disclosure of the program used and the specific portions drafted with it. These are individual orders, so read the standing orders for your judge rather than assuming.

At the system level, the picture is uneven. The Fifth Circuit proposed an AI certification rule in late 2023 and then declined to adopt it in June 2024, noting that saying you used AI would not excuse an otherwise sanctionable filing. Illinois adopted a Supreme Court policy effective January 2025 requiring review of AI-generated content for accuracy before submission. Florida went furthest: effective June 15, 2026, Rule of General Practice and Judicial Administration 2.515 requires the signer of any document filed in a Florida court to certify that all cited legal authorities exist and are accurately cited, and it binds self-represented litigants as well as lawyers. A parallel amendment to federal Rule 11 has been proposed to the Advisory Committee on Civil Rules but is not law.

The ethics baseline is settled regardless of jurisdiction. ABA Formal Opinion 512, issued in July 2024, holds that lawyers must understand a generative tool capabilities and limitations, specifically including its propensity to hallucinate, must independently verify its outputs, and that the duty of candor to the tribunal bars submitting hallucinated citations.