Catching a Hallucinated Citation Before It Undermined a $1.34 Million Demand

A fabricated liability clause. A nursing home negligence case in Las Vegas. A demand letter worth $1.34 million. Here's how Human AI Output Auditing caught a critical legal-research error before it reached the opposing side.

Case at a Glance

The Case

A Las Vegas-based personal injury firm was preparing a nursing home negligence demand on behalf of a family whose loved one suffered a serious injury while under the care of a skilled nursing facility. The case involved clear evidence of inadequate supervision and a documented failure to follow the facility's own care protocols — the kind of nursing home negligence claim where liability, if properly supported, is strong.

To move quickly, the firm used an AI drafting tool to generate the first version of the demand letter, citing the facility's contractual and statutory obligations alongside the medical chronology supporting the injury and damages. The draft demand totaled $1.34 million — a figure grounded in real, well-documented damages.

What the AI Got Wrong

Before the demand letter went out, the firm routed it through FactAssess' AI Output Audit service. During claim-by-claim source verification, our reviewers identified a serious problem: the AI-generated draft cited a specific liability clause — presented as part of the facility's licensing or contractual obligations — that did not actually exist in the underlying documentation or applicable Nevada nursing home regulations.

This is a textbook example of AI hallucination in legal drafting: language that reads as confident, specific, and authoritative, but that has no basis in the actual source record. Had this gone unverified, the demand letter would have rested part of its liability argument on a clause an insurance adjuster's counsel could disprove with a single phone call — undermining the credibility of an otherwise well-supported $1.34 million claim.

How the Audit Caught It

Our source-traceability review process is built specifically to catch this kind of error. Every factual and legal claim in an AI-generated demand letter is checked against the actual underlying documentation — medical records, facility policies, applicable regulations, and case-specific evidence — not accepted at face value because it reads convincingly.

In this case, when our reviewers attempted to trace the cited liability clause back to its source, no matching provision existed in the facility's documentation or in the relevant regulatory framework. The claim was flagged, verified as unsupported, and corrected — with the demand letter's legitimate, well-documented liability arguments (inadequate supervision, protocol failures, and the resulting injury) left fully intact and properly sourced.

The Outcome

FactAssess delivered an audited, corrected demand letter — same damages figure, same core liability theory, but with the fabricated clause removed and the argument rebuilt on claims that could withstand scrutiny. The firm was able to send a $1.34 million demand it could stand behind fully, backed by an audit trail memo documenting exactly what was reviewed and corrected.

Nursing home negligence and elder abuse claims routinely draw close scrutiny from insurance defense counsel, who actively look for exactly this kind of unsupported claim to discredit a demand. Catching the error before submission — rather than having it surface during negotiation or litigation — protected both the credibility of the claim and the firm's standing with the carrier.

Why This Matters for Every AI-Generated Demand Letter

This case illustrates exactly the risk our AI Output Audit service exists to catch: AI drafting tools are fast and often accurate on the facts they're given, but they can also generate confident-sounding legal language — citations, clauses, statutory references — that simply isn't real. In a routine soft-tissue claim, an error like this might go unnoticed. In a catastrophic nursing home negligence case worth over a million dollars, it's the kind of mistake that can hand the defense a credibility argument for free.

Independent, source-traceable human verification isn't a formality — it's what stands between a demand letter that reads well and one that holds up.

See How Our AI Output Audit Process Works → Request a Free Audit →

Don't let a fabricated citation undermine your next demand. Get it audited first.

Request a Sample Audit → See Our AI Output Audit Service →

Leave this field blank:

Submit