Check the brief you received for fabricated citations

What does a fabricated citation look like?

A lawyer already knows the pattern. The cite looks like the correct material around it: a plausible case name, a reporter that exists, a year that fits the argument. What fails is the record. The volume and page do not resolve to a decision. The case exists and does not say what the parenthetical claims. The regulation has been superseded, or was never the provision quoted.

Those errors are invisible on a read, including to the lawyer who filed them. That is why a second model's opinion is not a check. A chatbot can invent a confident confirmation the same way the first model invented the cite.

Why does receiving counsel get scolded for a cite they did not invent?

Because the court sees the chain, not the origin story. In a 2026 California appellate matter already described on this site, a fabricated citation began in a Reddit post, moved into a sworn declaration, then into an attorney's filings, then into opposing counsel's proposed order, and then under a judge's signature. Nobody in that chain invented it. Every one of them received it, and every one of them assumed someone upstream had checked.

The court sanctioned the attorney who introduced it $5,000 and referred the matter to the State Bar. Receiving counsel who copy a fabricated authority into a proposed order, or build a response on it, are in that same chain. The professional duty is to the filing you put in front of the court, not to the person who drafted the document you were served.

What did a pre-registered study actually measure, including what failed?

Figures dated March 2026 (pre-registration) and June 2026 (self-check of our own paper), as published on the validation-study page. No numbers are added here that do not appear there.

In March 2026 we pre-registered a validation study on the Open Science Framework, before collecting any data, with independent academic oversight. Twenty publicly traded U.S. companies across four analyst-coverage tiers. For each, a frontier model generated a detailed company profile from training knowledge alone. We then verified those profiles two ways: single-model self-check by three independent frontier models, and the full adversarial protocol.

Self-check error counts varied by up to 9.6:1 on the same document depending on which model did the checking. The harshest checker passed three companies' profiles as error-free; adversarial verification later identified 16 verifiable errors across those three documents.

In the study's second phase, the improved pipeline examined 1,829 claims across the twenty companies and surfaced 84 raw debunked claims, an estimated 65 to 75 genuine after accounting for documented false-positive categories. Most claims survive verification: 79.7% confirmed.

The precision figure above is an engineering estimate, not yet a human-verified number; independent human evaluation is in progress. One of our six pre-registered hypotheses failed on its registered metric, and we report it as registered rather than reframing it. One of our originally reported catches turned out to be a false positive in our own pipeline; we found it because we ran the Gauntlet on our own paper, and we corrected our numbers downward, cutting the Phase 1 catch count from 27 to 26.

Before releasing the study we submitted the paper itself to the Gauntlet: five drafts, the assembled manuscript, a run on author commentary, and a post-draft check in June 2026. Every run returned a failing grade. Eight documented failure modes are published in the paper, including entity-resolution failures, knowledge-cutoff false positives, phantom claim extraction, and cached verdicts replayed as fresh evidence.

How do you run the brief you received?

Open /analyze, upload the filing opposing counsel served, and wait for the Gauntlet Report. The engine checks citations against court and regulatory records, scores the claims with deterministic Bayesian math, runs a causal and argument pass, and returns a cryptographically signed certificate. It is not KYC, not an ID scan, not a plagiarism check, and not another model's opinion.

For the litigation desk this is built for, see /solutions/legal. For the pre-registered numbers quoted above, including what did not work, see /validation-study.