For whoever is on the receiving end
Last updated: August 2026 · this page changes when the system changes.
You get fifty decks a quarter and take every one of them on trust. Facticy reads the document you were sent and tells you what stands up, what does not, and what nobody outside the author could ever confirm — with the source behind each verdict.
The problem is not that decks lie
Most of what you receive is broadly accurate. The trouble is that you cannot tell which parts, and checking by hand costs an analyst a day per document — so nobody does it, and the numbers travel into a memo, then into a committee paper, then into a decision.
What you actually want is narrow: which figures in this document are checkable, which of those check out, and which ones nobody outside the author could confirm even in principle. That last group is not a failure — it is the part where you have to trust the counterparty, and knowing exactly where that line falls is most of the value.
What you get back
- Every claim, with a verdict and the sources behind it. True, false, partially true, disputed between models, or unverifiable — and if it is unverifiable, WHY: proprietary data, a forecast, a paywalled source, or a method the document never discloses.
- The document itself, annotated. Each finding marked on the page it came from, so a committee paper can cite a page rather than a summary.
- Which sources are independent and which are the author's own. A figure a company publishes about itself is the primary record for that figure, and it is not corroboration. The report distinguishes the two rather than counting them alike.
- Three scores: factual soundness, document quality, and a human-authorship estimate — each with the claims behind it, not a single opaque grade.
How it works in practice
Upload the PDF, or import it from Drive or OneDrive. There is nothing to install and nothing for the sender to do — they never learn the document was checked.
A single research pass builds one shared evidence base for the document; a panel of independent models then judges every claim against that evidence. Analysis takes minutes, not a day.
What it is not
- It is not a due-diligence provider. It checks what a document ASSERTS against public sources; it does not audit the business behind it, and it has no access to anything the counterparty has not published.
- It is not a lie detector. A claim that no public source can settle comes back as unverifiable, not as suspicious — most of them are simply internal data, and reading that as an accusation would be worse than not checking at all.
- It does not replace your judgement. It replaces the day an analyst spends assembling the evidence for it.
Who this is for
Investment committees reading pitch decks and market studies. Boards reading what management sends up. Audit and risk teams reading adviser reports. Anyone whose job includes the sentence "and where does that number come from?"