AI Label Checker

Choose an image and see every AI label its file carries: Content Credentials, the IPTC “made with AI” tag Pinterest reads, an AI maker’s name, a generator’s settings. Read on your device; nothing is uploaded.

This file says AI made itA sample: an image Google’s AI made, with the labels Google wrote into it. Choose yours.

Up to 30 at a time: JPEG, PNG, WebP, HEIC, AVIF or GIF.

    It reads the file’s hidden details, not the picture. A watermark hidden in the pixels, like Google’s SynthID, cannot be seen this way (below).

    What Pinterest says about AI labels

    • When our system detects via the metadata on a Pin that it's been generated or modified using AI, we add a Gen AI label to it.
    • To do this, we not only analyze a Pin’s metadata following the IPTC Metadata Standard, but also developed classifiers that automatically detect Gen AI content, even if the content doesn’t have obvious markers.
    • In addition, some Pins may also have the AI modified label if the content owner indicated that the Pin was generated or modified using AI.
    • If you find an AI label on one of your Pins and want to appeal it, contact our support team. Pinterest’s contact form, opened on its Gen AI label appeal
    • Pinterest’s help names the IPTC standard. It does not say whether it also reads Content Credentials.
    Pinterest Help: Gen AI labels, checked 4 October 2026.

    A watermark in the picture

    • SynthID adds an invisible digital watermark to an AI-generated image (or video segment). It’s added the moment content is created, and designed to stand up to modifications like cropping, adding filters, changing frame rates, or lossy compression. Google DeepMind: SynthID
    • To look for one, Google points to its Gemini app, which Uses invisible watermarks to determine if images, videos, and audio were generated or edited specifically by Google's AI models. While other companies have started to adopt SynthID watermarks, Gemini can currently only recognize content created by Google AI tools. Gemini Apps Help: Verify AI-generated images, videos, and audio
    Checked 5 October 2026.

    Where each label sits in a file

    LabelWhere the file keeps itWhat it says
    Content Credentials (C2PA)JPEG: the APP11 block · PNG: the caBX chunk · WebP: the C2PA chunk · HEIC, AVIF: a uuid box · GIF: the C2PA_GIF blockWho made the file and what was done to it, signed. An AI maker’s credentials say it with IPTC’s own words.
    IPTC digital source typeXMP, or inside the Content CredentialsHow the picture was made: “Created using Generative AI”, “Edited using Generative AI”, “Digital capture sampled from real life” and 17 more
    An AI maker’s nameEXIF Software, XMP CreatorTool, the credentialsThe program that made or saved the file
    Image-generator settingsPNG text: “parameters” (AUTOMATIC1111, Forge), “prompt” and “workflow” (ComfyUI)The prompt, the steps and the seed
    Words that say AIA credit, a caption or a comment“Made with Google AI”, “AI-generated”

    The C2PA Manifest Store shall be embedded as the data contained in an APP11 marker segment as defined in JPEG XT, ISO/IEC 18477-3. An AI maker’s credentials carry a c2pa.actions assertion starting with a c2pa.created action, itself having a value of http://cv.iptc.org/newscodes/digitalsourcetype/trainedAlgorithmicMedia in the corresponding digitalSourceType field.

    C2PA Technical Specification 2.2, Appendix A; IPTC NewsCodes: Digital Source Type (released 23 October 2024). Checked 5 October 2026.

    Quick answer: Choose one image or up to 30. For each file you see every AI label it carries, where in the file it sits, and what the file says overall: made with AI, edited with AI, it names an AI maker, Content Credentials with no AI label, a camera, or no label at all. A label is what the file says about itself, so it is a strong hint, not proof, and no label is not proof either. It reads the file, not the picture: a watermark hidden in the pixels, like Google’s SynthID, needs Google’s own check. Nothing is uploaded.

    How to use it

    1. Choose your images, or drop them on the tool on a computer. On a phone this opens your photo library. They stay on your device.
    2. Read each card. The top line is what the file says overall. Each row is one label, with where the file keeps it. “Also inside” lists other hidden details: location, camera, date.
    3. Before you pin it, the first section below quotes what Pinterest says it does with these labels.

    The labels it reads

    • Content Credentials (C2PA). A signed record of who made a file and what was done to it, written by the program that saved it: some AI image makers write them, and so do some cameras and editing apps. The card shows the program the credentials name.
    • IPTC digital source type. One short code for how the picture was made, from the IPTC’s list of 20. “Created using Generative AI” and “Edited using Generative AI” are the AI ones; “Digital capture sampled from real life” is a camera. It sits in the file’s XMP, or inside its Content Credentials. This is the standard Pinterest’s help names.
    • An AI maker’s name, where a file names the program that made it (EXIF Software, XMP CreatorTool).
    • Image-generator settings. Stable Diffusion apps such as AUTOMATIC1111 and Forge save the prompt, steps and seed in a PNG; ComfyUI saves its whole workflow.
    • Words that say AI in a credit, caption or comment, such as “Made with Google AI”.

    IPTC’s digital source types

    IPTC’s own name for each code, and how this checker sums it up. The list is IPTC’s, released 23 October 2024.

    CodeIPTC’s nameThe checker says
    trainedAlgorithmicMediaCreated using Generative AIAI made it
    compositeWithTrainedAlgorithmicMediaEdited using Generative AIAI changed part of it
    compositeSyntheticComposite including generative AI elementsAI changed part of it
    compositeComposite of elementsa mix that may include AI
    virtualRecordingVirtual event recordinga mix that may include AI
    digitalCaptureDigital capture sampled from real lifecaptured from real life
    computationalCaptureMulti-frame computational capture sampled from real lifecaptured from real life
    negativeFilmDigitised from a transparent negativecaptured from real life
    positiveFilmDigitised from a transparent positivecaptured from real life
    printDigitised from a non-transparent mediumcaptured from real life

    The other ten (human edits, a screen capture, data-driven media, a picture made by a formula and so on) are shown on the card by IPTC’s name and leave the overall answer as it is.

    FAQ

    How can I tell if an image is AI-generated?
    Start with its file: choose it above. An AI label (Content Credentials, the IPTC “Created using Generative AI” tag, a generator’s settings) is the clearest sign there is. If there is none, the file may have been saved again without its hidden details, so look at the picture itself too, and for an image from Google’s AI, ask Google’s Gemini app to look for its SynthID watermark. No single check is proof.
    Does Pinterest label AI images?
    Yes. Pinterest’s help says it adds a Gen AI label when a Pin’s metadata says it was generated or modified with AI, that it reads that metadata following the IPTC standard, and that it also uses its own classifiers to find AI content without obvious markers. A Pin can also be labelled when its owner says it was made or changed with AI. Its words are quoted above.
    What are Content Credentials, or C2PA?
    A standard record of where a file came from and what was done to it, signed by the program that wrote it. The C2PA (the Coalition for Content Provenance and Authenticity) writes the standard and calls the records Content Credentials. This checker shows that a file has them and which program they name. It does not check the signature.
    Why does my AI image show no label?
    Not every AI image maker writes one, and a copy made by a screenshot, a messaging app, a website or an editor often leaves the hidden details out. No label means the file says nothing about how it was made, not that a person or a camera made it.
    Does One-Hour Photo remove these labels?
    Yes, with every other hidden detail: in One-Hour Photo, with Remove hidden details ticked, as it is when the page opens, the photo it saves carries none of them. A watermark hidden in the pixels is part of the picture, and Google says SynthID is designed to stand up to filters. If a picture is AI-made, saying so is up to you; Pinterest’s help says a Pin can carry its AI modified label when its owner says it was made or changed with AI.
    Are my images uploaded?
    No. Each file is read by your browser, on your device. The link this page makes holds nothing about your images.