Is this real? Here’s how we’d know.

ASURIQ doesn’t ask another AI whether the first one was right. It checks claims against the same kind of sources a researcher or a compliance officer would reach for, weighs how independent that evidence actually is, and turns the result into a permanent record you can click into. This is that process, in plain terms — the formal version lives on the methodology page.

Extract, check, weight, certify, recheck.

We never see the page.

Reading the page and pulling out checkable claims happens on your own AI — the same one you’re already talking to — not on ASURIQ’s servers. Only the extracted claim goes out for a lookup. This isn’t a policy promise; it’s how the extension is built. “We never see what you browse” is architecturally true, not just stated.

Real databases, routed by domain.

A medical claim checks PubMed, FDA, ClinicalTrials.gov, and WHO. A financial claim checks SEC EDGAR, FRED, and BLS. A legislative claim checks Congress.gov. Twenty-one databases free on every request; three commercial sources — OpenCorporates, OpenSanctions, Nominatim/OSM — add on the detail card for ownership and sanctions depth. Full list on the methodology page.

Three scores, six evidence axes.

Formal definitions, the Admiralty Code lineage, and what these axes deliberately don’t measure: see the methodology page.

Four sources. Three real confirmations.

Agreement isn’t the same as corroboration if the sources share an owner. Here is a real ASURIQ certificate showing the difference.

Every verdict resolves to a record.

The evidence table, the independence readout, the three scoring categories, and the verdict stated in words — permanent, linkable, and readable by a machine as well as a person. A green square is easy to fake in a screenshot. A certificate that actually resolves at a stable URL is not.

Verified as of, not verified forever.

Every certificate states when it was checked and when it will next be rechecked. Slow-moving facts — a regulatory filing, a scientific consensus — get rechecked on a long cycle. Breaking news gets a “Pending” state instead of a rushed guess: sources haven’t caught up yet, and a verification authority doesn’t race to be first. That buffer is a quality guarantee, not a gap.

Four ways a check comes back empty — named, not hedged.

“We cannot verify this from public sources” is not the same claim as “this is unverifiable,” and neither is the same claim as “this is false.” Silence would look like withholding; naming the reason shows the work.

No appeals process. Just better evidence.

There’s no editorial judgment behind a verdict to argue with — only the arithmetic. Submit public data supporting the alternative and the score re-runs. It moves if the evidence does; it doesn’t if the evidence doesn’t. A dedicated submission flow is coming; for now this reaches a person directly.