How PerilAI decides what’s real.
Every record is a real, documented matter, filed by the insurance line it hits, carrying its sources and an honest confidence label. Here is exactly how a claim gets into the database, and how we keep it trustworthy enough to underwrite against.
Where the data comes from
We draw from primary public records first, and reputable secondary reporting only to fill gaps, never to stand alone where a primary exists.
Every record carries its citations, each tagged primary or secondary. You can audit all of them on the Sources page, which also flags any citation that points at a search page instead of a specific document.
Every record carries a verification tier
Confidence is a label on the record, not a hidden judgment. It reflects how well the legitimacy and the details of the matter are sourced.
A human reviews before anything publishes
Nothing is auto-published. Candidates surfaced by our weekly sweep of public sources are screened, then a person confirms the facts and clicks through every primary citation before a record goes live. That review gate is the difference between a database you can underwrite against and a scraped list.
Honest labeling, two more ways
The underwriting read
Beyond the facts, most records carry a short coverage read from the PerilAI desk: which towers the fact pattern would touch, where the coverage fight is, and what the exclusion landscape looks like. That is the layer that turns a lawsuit into an underwriting signal.