Pinterest's "AI modified" label, and what removing metadata can't do

Pinterest’s “AI modified” label comes partly from a Pin’s metadata. Its Help page says it analyses metadata “following the IPTC Metadata Standard”, and removing that metadata from your copy of an image, confirmed by a re-scan, takes the signal away. Pinterest also says it has “developed classifiers that automatically detect Gen AI content, even if the content doesn’t have obvious markers”, and no file cleaner changes a classifier. Cleaning also does nothing for a Pin that is already live.

Our prediction: high confidence, for the label that metadata triggers. We expect an image cleaned with NoWatermark using Everything removable (not Provenance-light, which keeps IPTC), with the re-scan showing IPTC and XMP as removed, not to receive the label Pinterest applies from IPTC metadata. Pinterest’s classifiers are separate: they can still label an image that carries no metadata, and we cannot predict what they will decide. This is a prediction from Pinterest’s documentation, not yet our own test result. We are running that test and will publish the result on this page.

Try it: check and clean your image before you post. It runs in your browser, and your file is never uploaded to us.

What the label is and where it shows

If Pinterest’s system detects that AI was used to generate or modify a Pin that is not in a promoted state, you see an “AI modified” label in close-up, at the bottom left, after you click on the Pin. Pinterest announced the label on 30 April 2025, describing it for image Pins.

The wording covers both cases. The label says “AI modified” whether Pinterest’s system detects that AI was used to generate a Pin or to modify it. So a real photograph that went through an AI-assisted edit can carry the same label as a fully generated image.

For ads, the label does not appear when you click on the Pin. Pinterest says you can see it along with other details about the ad by clicking or tapping the ellipsis icon next to the Pin.

What triggers the label on Pinterest

Platforms add labels from four sources: metadata read at upload, invisible watermarks, visual classifiers and self-disclosure. Pinterest documents three of them:

• Metadata. Pinterest names the IPTC Metadata Standard. It does not say which IPTC fields it reads or which values count. The IPTC vocabulary defines trainedAlgorithmicMedia as “Created using Generative AI” and compositeWithTrainedAlgorithmicMedia as “Edited using Generative AI”, but Pinterest does not say it acts on either. Google said in October 2024 that photos edited with tools like Magic Editor, Magic Eraser and Zoom Enhance already include IPTC metadata indicating generative AI editing. That shows such editors write this kind of signal. It does not show what Pinterest does with it.
• Classifiers. This is what sets Pinterest apart in the sources we relied on. Pinterest says its classifiers detect Gen AI content “even if the content doesn’t have obvious markers”. We cannot measure or influence a classifier from the file.
• Self-disclosure. Pinterest says some Pins “may also have the AI modified label if the content owner indicated that the Pin was generated or modified using AI”.

Pinterest’s Help page does not mention C2PA. We do not know whether Pinterest reads it, and this page does not claim it does or does not.

What removing metadata changes, verified

NoWatermark scans an image for C2PA manifests, XMP, IPTC and AI generator metadata. For XMP it shows a DigitalSourceType value as “Created by a generative AI model” or “Composite including AI-generated elements”. It removes what you choose, scans the cleaned file again, and marks a signal as removed only if the second scan cannot find it. The image data is copied byte for byte, not re-encoded.

Because Pinterest names IPTC, the choice of preset matters in the Remove AI label tool:

• Provenance-light removes Content Credentials, XMP and embedded text and keeps EXIF and IPTC. On Pinterest, that leaves the metadata standard Pinterest names in place.
• Everything removable removes EXIF, IPTC, XMP, Content Credentials and embedded text. One EXIF rotation field is kept by default.
• Privacy-safe removes location and device data and keeps provenance.

Even “Everything removable” leaves classifiers untouched. We have not yet tested a cleaned file on Pinterest, so we cannot tell you what label, if any, it would get. Our scanner also does not read the DigitalSourceType value inside a C2PA manifest, so it cannot tell you whether a credential says generated or edited.

What it cannot change

SynthID and other invisible watermarks are always reported as Unable to verify. NoWatermark cannot detect them, cannot remove them and cannot confirm they are absent. See our SynthID guide.

Visual classifiers are the same kind of limit. They are not something a file cleaner can measure, and Pinterest is the platform in our sources that says it uses them for these labels. If a label is wrong, the route Pinterest documents is the appeal below, not a change to the file.

If your Pin is already labelled

Cleaning changes the file on your device, not a Pin that is already live. Re-uploading a cleaned file creates a new Pin.

• Appeal. Pinterest’s Help page says: “If you find an AI label on one of your Pins and want to appeal it, contact our support team.” It names no form, criteria or turnaround, and we have not found them.
• False positives. Pinterest’s newsroom post says: “we know these systems aren’t perfect. This is why we set up an appeals system for creators and users who believe that their content may be mislabeled.” TechCrunch, reporting on the launch the same day, said creators will be able to appeal to Pinterest if their content is labelled incorrectly.
• Self-labelling. Pinterest describes a label applied when the content owner indicates the Pin was generated or modified using AI. Its help page does not say where a creator sets this, and we have not found it.

If a real photo was labelled after an edit in an AI-assisted tool, see why is my photo labelled AI for how those editors write metadata, and what is IPTC DigitalSourceType for the field involved.

Pinterest’s disclosure rules

The sources we relied on describe Pinterest’s labelling, not a requirement to disclose. Pinterest’s Help page says a label may appear where the content owner indicated AI use, and that Pinterest’s own system may add the label. We did not find a Pinterest statement in those pages that requires creators to disclose AI-generated images, and we have not reviewed Pinterest’s wider community rules.

For advertisers, the Help page says the label is shown in the ad details, not on the Pin itself (see above).

Cleaning a file does not change what a platform expects of you. If an image is AI-generated or AI-modified, say so where Pinterest lets you.

Related

For other platforms, see the guides on Instagram, X, TikTok photo posts and LinkedIn.

Try it on your own image

Drop your image into Remove AI Label Metadata. It lists every signal the file carries, removes the ones you choose, and scans the cleaned copy again so you can see what is gone and what remains, including anything it can only mark Unable to verify. It all runs in your browser: your file never leaves your device. Use Everything removable for Pinterest, since Provenance-light leaves IPTC in place.