How we check

Sources

Instructions for specific products, such as iPhone or Microsoft Word, come from the maker's own support pages. Facts about file formats come from the published standard or the standards body. Every source is listed at the end of the page that uses it, with the date we last checked it and whether we read the full text or only part of it.

We label each source by type: official documentation, standard, standard vocabulary, or official tool. Where we only confirmed a document's title and publication details, we say so.

What we don't do

• We don't publish claims about how social networks or messaging apps handle metadata unless we have tested them ourselves and can show the date and method. We haven't done that yet.
• We don't use invented statistics, testimonials or reviews.
• We don't claim the tool detects AI-generated images. It reports labels that are in the file, which can be missing or false.

How the tool is tested

Automated tests create sample files with known, made-up metadata (a fictional name, a camera serial number, a GPS position, AI prompt text, a PDF with an earlier saved version, a Word file with comments and tracked changes). For each format the tests check that the findings are reported, that cleaning removes them, that the made-up values no longer appear anywhere in the cleaned file's bytes, and, for JPEG and PNG, that the image data is unchanged byte for byte. These tests run before every release.

Risk ratings

High, medium, low and info ratings are editorial judgments about how much a field can reveal on its own. They are explained on the field reference. They are not a standard and do not replace your own judgment about your situation.

Updates and corrections

Guides show when they were published and when they were last substantively updated, with an update history. Small fixes such as typos don't change the date. To report a mistake, email info@metadatachecker.com; confirmed errors are corrected and noted in the page's update history.