How to check if a photo has AI metadata

Two checks, in this order

First, tap AI info on the live post or Story if one exists. Meta uses that screen to say whether the label came from industry-shared signals or from a person disclosing it. Second, Inspect the file you believe you uploaded. Those answers can disagree, and the disagreement is the finding: a chip that cites industry signals plus an empty file report means you scanned a different file, or the signal is not metadata this parser lists (including a pixel watermark).

Photoshop and Lightroom exports are a common mismatch. The project can save with Content Credentials on even when a later “save for web” dropped them, or the reverse. Inspect the exact export. Try sample is only a demo of the report layout. It is not your photo.

Inspect before you strip

The useful question is not “is this AI?”. The useful question is “what is written next to the pixels?”. Camera apps write EXIF. Phones often write GPS. Generators and some editors write C2PA Content Credentials, XMP that points at those credentials, or IPTC DigitalSourceType. Automatic1111-style PNG exports often write a parameters chunk — that recipe has its own page, Stable Diffusion PNG metadata. A C2PA-only question is check if an image has C2PA credentials. Any of those can be listed without sending the image to a server.

Open the AI metadata remover and drop the still you might post. You can stop after the before report. Pressing Clean this file is optional and only useful after you have seen a category you want gone. Generation tags have their own page, the AI metadata cleaner. The comparison with AI Metadata Cleaner is the alternative page. JPEG, PNG, and WebP are read directly. HEIC, AVIF, GIF, TIFF, and BMP are read as well, and the report says which format you get back.

What the field report is telling you

A long report is not a verdict that Instagram will label the upload. It is an inventory. A short report is not a verdict that Instagram will stay quiet. Instagram can label from signals that never appear in this inventory. That limit is the subject of Remove Instagram AI Info.

Run the check

A second look, still without a server

Content Credentials can also be inspected with a public verifier that understands C2PA manifests. That kind of check answers “is there a manifest?”, which overlaps the C2PA chip here. It still does not answer “will Instagram show a badge?”. Use one or both inspections on the file. Do not confuse either result with a classifier score.

Phone galleries often show a slice of EXIF and hide manifests. “No info” in the iOS or Android info sheet is not a C2PA test. The browser tool on this page is aimed at the categories social apps have started to read, plus the GPS fields people mean when they say a photo is leaking location.

When a clean report still leads to a label

You can do this check correctly and still see AI info after upload. Causes that fit a clean report:

• The image was edited with Instagram’s own AI tools after the check.
• You uploaded a different file than the one you inspected (a cloud original, a second export, an uncleaned duplicate).
• Instagram matched or classified the pixels without needing metadata.
• A pixel watermark is present. This scanner does not display SynthID or TrustMark.

Those cases, including a Story reshare and a carousel slide that was not cleaned, are on Remove Instagram AI Info.

Want a general metadata viewer rather than an AI-focused check? View image metadata online shows the GPS, camera, XMP, IPTC, and C2PA rows for any image, with the same inspect-only mode and no upload.

Check photo AI metadata in the inspect report: C2PA, AI XMP, and GPS show up when they are in the file. An empty report is a result, not a promise the picture has no watermark. Check photo AI metadata again on the After list if you clean the file.

If the question is whether the file still carries generator paperwork, use check if image is AI. That page reads metadata only and says what it will not decide about the pixels.