Verticals & use cases

One architecture, every registry

Swap the registry — the gate does the rest.

The only domain-specific piece is the source of record. Two verticals are proven; three more are in active build; the rest are a registry away.

⚖️ Legal

Proven

CourtListener

Fabricated & mis-cited case law — the Mata mode. 366 cites, frontier 7–56% bad, gate 0.

We verify a citation exists, is correctly identified and is accurately quoted — not that it supports the proposition it is cited for. Scope.

📊 Finance

Proven

SEC EDGAR / XBRL · or your own model

Figures that don't match the filing — or don't match your own model, where no public registry exists. 5,000+ claims, frontier 33–67% on dirty data, gate 0/6,000.

🧬 Medical / Scientific

Proven

PubMed · ClinicalTrials.gov · FDA

Fabricated PMIDs & trial ids; real id, wrong finding; invented studies.

🎓 Academic

Proven

Crossref DOI · OpenAlex

Fabricated DOIs; real DOI attributed to the wrong paper.

🛡️ Software / Security

Building

NVD / CVE · PyPI · npm

Hallucinated CVE ids and invented packages ("slopsquatting") — a distinct error mode.

📑 Patent · Regulatory · Gov data

USPTO · FDA · Federal Register · Census

Fabricated patent/docket numbers, wrong approval status, mis-stated official figures.

Proven — full finance/legal-style benchmark Building — verifier + calibration underway Next — a registry integration away

A new vertical is a new source of record — not a new product. The gate, the model-independence, and the compounding memory are shared across every one of them, and each is held to the same standard before we call it proven.

Two attorneys presenting argument in a courtroom — Legal — every citation resolved against CourtListener.

Works in every vertical · New

When there is no public registry: check it against your own documents

A citation can be checked against the court’s docket. A portfolio company’s capacity factor cannot — no registry holds it. So upload the report and the model, accounts or exhibits behind it, and we reconcile the figures against your own records: what agrees, what disagrees (quoted against the source row), and what we could not check and why.

Corroboration against your own documents — a narrower claim than verification against an external register, and the certificate says so. From $5.50, capped at $9. Pricing · Try it

Built for accuracy-critical work

Where a single wrong fact is expensive.

Strongest where an authoritative registry exists — but valuable anywhere accuracy matters and facts can be checked.

⚖️

Legal briefs & memos

Every citation resolved to the real case. Catches the Mata-class error before a judge does.

Registry: CourtListener

📊

Financial reports & models

Every reported figure checked against the filing of record. A wrong number can't propagate into the valuation.

Registry: SEC EDGAR / XBRL

🔍

Due-diligence memos

Claims, figures, and citations verified end-to-end before they reach the deal team.

Multi-registry

🧬

Medical & scientific writing

Every study, statistic, and identifier checked. Fabricated or mis-attributed references flagged.

Registry: PubMed / FDA

📋

Regulatory filings

Facts and figures reconciled to source before submission, with a full audit trail.

Multi-registry

🎓

Grant applications

Cited literature and reported data verified — reviewers won't find an error you missed.

Registry: PubMed

✍️

Essays & college assignments

Every quote, statistic, date, and citation fact-checked. We verify the facts — we don't write or grade the argument.

Fact & citation check

✅

Anything accuracy-critical

If it contains claims that can be checked against a source of record, AttestAlly can verify it.

Ask us

An analyst presenting growth figures to a boardroom — Finance — every reported figure reconciled to the filing of record.

Private markets

No ticker? No filing? We can still check the arithmetic.

A deal memo, fund letter or board pack usually has nothing an external registry can resolve — no ticker, no filing reference, no reporter citation. Most verification tools stop there. We check the document against itself.

Totals against their parts

Segment figures that don’t add to the stated total, and balance-sheet identities that don’t hold.

Percentages against their bases

A share stated as a percentage of a number that is also in the document — and doesn’t reconcile.

↗

Growth against its endpoints

A stated change that the two figures either side of it do not produce.

What this is, and what it is not. This is internal consistency: the document tested against its own stated figures, with no external source involved. A memo can be perfectly self-consistent and still be wrong about the world — internal arithmetic cannot tell you a market size is overstated, only that your own numbers disagree with each other. Where a document carries no external anchor, the certificate says so plainly rather than letting the result read as corroborated verification.

Measured on our benchmark corpus: 0 false accusations across 107 registry-free documents that were correct, and 89.1% of documents carrying a planted arithmetic error were caught. Those figures are ours, from our own corpus — treat them as indicative.