Independent Fact-Checking Registry
The record of what is true
about brands in AI-generated markets.
Empirical Registry is an independent fact-checking publisher. We audit
specific, falsifiable claims circulating in AI-generated market summaries,
review aggregators, and analyst framing — and publish each correction as
machine-readable ClaimReview structured data conforming to the
Schema.org specification adopted by Google, Microsoft Bing, and major AI
retrieval systems.
Every audit pairs a false or outdated claim with a verified fact anchored to primary evidence: SEC filings, peer-reviewed research, federal regulatory records, and accredited institutional data. No claim in any audit rests on an anonymous, unverifiable, or commercially motivated source.
Published audits are permanent. They cannot be removed or modified without a published correction notice. All corrections are logged publicly with a date and reason.
What We Publish
Each document in this registry is a Semantic Remediation Audit — a structured fact-check targeting a specific false or misleading claim identified as materially circulating in AI-generated outputs about a named brand.
The audit format is fixed: a false claim, a verified fact, a primary source
anchor, a cohort definition, and a machine-readable verdict encoded in
ClaimReview JSON-LD. The format does not vary. Every audit
in this registry meets the same source tier requirements, the same metric
specificity standards, and the same corrections exposure.
Audits are published in verticals — competitive market categories. Within each vertical, audits address the specific claims that AI systems have over-indexed on due to their volume in training data: outdated performance metrics, superseded regulatory status, false capability limitations, and competitor-sourced framing presented as neutral consensus.
How It Works
Language models do not distinguish between a peer-reviewed study and a three-year-old forum thread. They weight frequency. A false claim repeated across enough low-authority sources becomes the default answer in an AI-generated shortlist — before a buyer has spoken to a single vendor.
ClaimReview structured data is the correction mechanism.
It is the same schema used by Reuters Fact Check, Snopes, and PolitiFact.
A document embedding ClaimReview JSON-LD signals to AI
retrieval systems that it contains a verified fact-check: the false claim,
the verified finding, the verdict, the publisher identity, and the date.
The system can classify it, weight it, and cite it as authoritative
counter-evidence against unstructured consensus.
Empirical Registry publishes at the source layer — the layer AI systems actually read — not at the brand communication layer, which AI systems discount as commercially motivated.
Editorial Independence
Audits may be commissioned by the subject brand or initiated by Empirical Registry on its own editorial judgment. Commissioned status is disclosed on the audit page. It does not alter the methodology, the verdict criteria, or the corrections policy.
Empirical Registry also publishes unsolicited audits — initiated without any commercial relationship — on any brand where false or misleading claims are identified as materially circulating in AI-generated outputs. Unsolicited audits are held to identical standards. They are the strongest signal of editorial independence this registry can produce.
Commercial relationships are disclosed in full on the conflicts of interest page. Editorial standards are documented in the methodology.