EXIF & Provenance Viewer
Free · No sign-up · Runs in your browser — your file never uploads.See what your file says about itself before the platforms do.
Drop a file here or click to choose — the file never leaves your device either way.
Verified by its actual bytes, read locally, forgotten when you close the tab.
What this viewer reads
- EXIF — camera make/model, software, capture date, and a clear warning if GPS coordinates are embedded.
- Provenance markers — C2PA Content Credentials presence, IPTC DigitalSourceType (the field that declares AI generation), and known generator/software strings.
- PNG chunks & XMP — text chunks (including AI-tool "parameters") and XMP packets, shown as readable evidence.
- An honest summary — evidence found, worded as evidence. Never a fake "87% AI" score.
What this tool cannot tell you
- No metadata ≠ not AI. Metadata is stripped by most apps and platforms; a clean file proves nothing either way.
- SynthID cannot be read here — it lives in the pixels and only Google's own portal can check it. We say so instead of pretending.
- A C2PA manifest proves a named signer asserted the file's history — read the signer, don't treat the badge as truth itself.
Why provenance is suddenly a creator problem
Platforms now label content automatically when they find machine-readable AI markers, and the EU's AI Act (Article 50, enforceable since August 2026) makes machine-readable marking a legal duty for AI providers — with deployer duties phasing in to December 2026. That means the metadata inside your file increasingly decides whether your post carries an "AI" label, wherever in the world you upload from. Reading that metadata before you publish — instead of being surprised after — is the whole point of this viewer.
A pre-upload provenance routine
Provenance and privacy pull in opposite directions, so decide which one you are optimising for before you export. If you may ever need to prove a photo or video is yours and real — for a platform appeal, a licensing dispute, a newsroom sale — you want the chain intact: camera originals untouched, C2PA credentials preserved, edits made on copies. If you are sharing everyday material publicly, the priority flips: the same GPS coordinates that prove where you were also tell strangers where you live.
Export settings decide which of those futures you get. Most editors have a metadata option buried in the export dialog ("include all metadata", "embed Content Credentials", or a strip-everything preset), and its default differs per app and per update. Every re-encode is a rewrite: the new file only carries the metadata the exporting tool chose to copy, and tools differ — one editor preserves EXIF but drops XMP, another keeps a C2PA manifest but flattens its history into a single "edited" claim. A screenshot is the extreme case: a brand new file with none of the original evidence at all. Platforms re-encode too; most strip EXIF on upload, but several read the provenance markers first and act on them, which is why the file you upload matters more than the file the audience eventually sees.
The routine itself takes a minute. Keep the untouched camera originals and project files on a drive that is not part of your publishing pipeline. Export a separate deliverable for each platform instead of reusing one file everywhere. Run the deliverable through this viewer before uploading: confirm the markers you expect are present (or absent), and check the GPS flag before you share any original file directly by message, email or forum. Sixty seconds of checking beats reconstructing your evidence after a label or a dispute arrives.
Frequently asked questions
What does this viewer actually read?
The metadata inside your file: EXIF fields (camera, software, dates, GPS presence), PNG text chunks, XMP packets, IPTC source-type values, and whether a C2PA Content Credentials manifest is embedded. All of it is read in your browser — the file never uploads.
If no AI markers are found, is the image definitely real?
No. Absence of metadata proves nothing — most editors strip it, and pixel-level watermarks (like Google's SynthID) are not stored as readable metadata at all. A clean scan means 'no machine-readable evidence found', never 'proven human-made'. Any tool that tells you otherwise is overstating.
Can this detect SynthID?
No tool outside Google can. SynthID is embedded in the pixels, not the metadata, and Google offers no public detection API. Google's own SynthID Detector portal is the only route, with small daily limits. What we CAN read are C2PA manifests and IPTC source types, which many AI tools and cameras now embed.
What is a C2PA manifest?
Content Credentials: a signed packet some cameras and AI generators embed, stating who made the file and what happened to it. Finding one is real evidence. This viewer reports its presence and the claims it can read; full cryptographic validation (including how much of the file the signature actually covers) is a deeper check we label separately.
Why does the GPS warning matter?
Many photos carry exact GPS coordinates in EXIF. Platforms strip them on upload — usually — but the file you share directly (messages, email, forums) may carry your home location. This viewer flags GPS presence so you know before you share.
Do you keep my file or its metadata?
No. Everything runs in this tab. Close it and the data is gone; nothing is sent to or stored on any server.
Related tools
- AI Label Checker — what will each platform do with this file?
- Thumbnail Checker & Preview — check the image's job performance too.
- Privacy Scan — before you share a photo or video, check what it reveals about you (faces, location, metadata).
Related guides
- AI labels in 2026: YouTube, TikTok and Instagram — what each platform requires you to disclose, and how provenance data feeds it.
CreatorChecks is not affiliated with YouTube, Google, TikTok or Meta.
Was this result accurate?
Thanks — feedback like this helps us catch errors.
Could not send — please try again.