FreeInformation, not adviceEvery page, or it says so

Prove every page was reached — not just the ones it skimmed.

Cross-checking models against each other catches one that hallucinates. It does not catch a pipeline that under-reads — quietly skimming past a page that doesn’t retrieve well, while every tool agrees. Upload a PDF and TrustAI partitions it, reaches every page, commits each page’s text to a hash, Merkle-roots them, and Knox-anchors a tamper-evident Reading-Coverage Receipt — court-checkable proof that every page was reached and committed, nothing skimmed past.

Why this exists. Top-k retrieval (RAG) is a skim: it returns the chunks that match a query and discards the rest. A clause that embeds poorly is simply never seen — and no amount of model cross-checking surfaces a page none of them read. A coverage receipt is the opposite guarantee: the document is the union of its pages, each reached exactly once, proven by construction, not by trust.

Your document

Choose a PDF to check (free tier; no document content is stored — only its hashes and the coverage receipt)

Coverage receipt

Your receipt will appear here: how many pages were reached and read, which (if any) need image-reading, the Merkle root, and the Knox anchor — one click from the per-page detail.

What the receipt proves — and what it never claims

Every page reached. The file is partitioned into disjoint, gap-free page spans; a deterministic gate proves pages 1…N are each covered exactly once. A missing or duplicated page fails the gate — the document is never certified as “fully read” with a page unaccounted for.
Each page committed. Every page’s extracted text is hashed; the page hashes roll into one Merkle root. Change one character of one page and the root changes — re-running verify names the exact page that moved.
No page silently skipped. A page with no text layer (scanned or image-only) is flagged for image-reading, never passed over — under-reading is made visible, not hidden.
It is not a legal conclusion. The receipt proves every page was reached, its text committed, and the whole is unaltered. It does not interpret the document or opine on your matter — that is your attorney’s judgment.

What it proves, precisely

This checker reads and commits the text layer of every page and flags any page without one. The receipt proves coverage and integrity — that every page was reached, its text committed to a hash, and the whole sealed under one tamper-evident root. It does not read the document for meaning or interpret your matter; reading for meaning, and every legal judgment, remain your attorney’s.

How the proof holds up

Property	How it is established
Completeness (no page skipped)	Deterministic partition + coverage gate — pages 1…N each in exactly one span, re-derivable from the file
Integrity (nothing altered)	Per-page SHA-256 → Merkle root; `verify` re-ingests the original and names any page-level mismatch
Admissibility (provable later)	The root is ML-DSA-signed and hash-chained on the TrustAI Knox chain when the receipt is created; the chain’s Merkle root is submitted to OpenTimestamps for Bitcoin anchoring on the chain’s anchoring cadence — FRE 902(13)/(14) audit posture

TrustAI is not a law firm and does not provide legal advice. This tool proves every page of a document was reached and committed and that it is unaltered; it does not interpret the document or tell you what to do. Those judgments belong to a licensed attorney. What TrustAI does, and never does →