Documented behavior only · every cell sourced + dated
Will your disclosure survive the upload?
Embedded disclosure only works if the platform reads it. Drop your file (or tick its signals), pick a platform, and see the documented outcome for each signal — built exclusively from what platforms have published about their own pipelines. Where a platform publishes nothing, this tool shows the gap instead of guessing.
Drop a file to read its signals
Or skip the file and tick the signals below by hand. The file is read locally — never uploaded.
The honest limits of a simulator
A platform's published statement describes the behavior it chose to document — typically, which metadata it reads to apply its AI label. It does not describe everything its pipeline does to your bytes, and pipelines change without changelogs. So this tool answers one question precisely — is there a documented path for my disclosure on this platform? — and refuses the other one. For what actually happened to a real upload, diff the real files with the Strip Report.
Questions
What does “documented” mean in this simulator?
It means the platform itself has published the behavior — in its help pages, newsroom, or policy documents — and the simulator links to that publication and stamps the date it was checked. If a platform publishes nothing about a signal, the cell says “not documented.” The simulator never upgrades a rumor, a forum post, or a third-party test into a platform claim.
Which signals do platforms actually read?
On the published record: YouTube says it may automatically label content that contains C2PA metadata; TikTok says it reads Content Credentials (C2PA) to instantly recognize and label AI-generated content from other platforms; Meta says it labels content when it detects industry-standard indicators, naming C2PA and IPTC metadata — and the IPTC digital source type lives inside the XMP packet. X publishes no provenance-reading program we could source.
If a platform reads my C2PA manifest, am I done disclosing?
No. The platform's automatic label is one layer. The legal duties — the EU AI Act's visible disclosure, China's explicit label, a platform's own creator-disclosure rule — still sit with you, and a manifest can be stripped by any re-save between you and the platform. The robust stack is layered: embedded provenance, a visible label in the creative, caption wording, and the platform's upload-time disclosure toggle.
Why not just test real uploads instead?
You can — that is exactly what the Strip Report on this site is for: it diffs your master against the copy a platform serves back. This simulator is the before-you-post complement, projecting only what platforms have put in writing. Measurements beat projections, and the page says so; use the simulator to plan, and the Strip Report to verify.