What the Instagram AI label is

In 2024 Meta began labeling images on Instagram, Facebook and Threads when it detected “industry standard AI image indicators”. The first label read Made with AI. After photographers complained that lightly retouched real photos were being tagged, Meta renamed it AI info in July 2024 and, from September 2024, moved the label into the post menu for content that was merely edited with AI tools rather than generated.

The label still appears prominently when the file says the whole image was AI-generated, and it is applied automatically at upload time. Once a post is up, the label cannot be removed by editing the caption. You have to remove the trigger before uploading.

What triggers it

  • IPTC digital source type. An XMP field that reads trainedAlgorithmicMedia (fully generated) or compositeWithTrainedAlgorithmicMedia (partly generated, such as Generative Fill). Adobe, Google, Microsoft and Midjourney write it.
  • C2PA Content Credentials. A signed manifest listing the software and actions. DALL-E, Adobe Firefly, Photoshop, Microsoft Designer and several cameras embed it.
  • Invisible watermarks. Meta has said it works with partners whose generators embed invisible markers. These live in the pixels and survive metadata removal.
  • Self-disclosure. The toggle you set when posting. Meta requires it for realistic AI video and audio.

Remove the trigger before uploading

  1. Check the file

    Drop the image into the tool below. It reads the C2PA manifest, IPTC digital source type and generator traces and tells you which signal Instagram would act on.

  2. Strip or re-encode

    Strip metadata keeps the pixels untouched. Re-encode writes a fresh JPEG or WebP with an optional 3% shrink when you also want new pixels.

  3. Upload the clean copy

    Download the cleaned file and post that one. Do not re-open it in an editor that writes Content Credentials, or the record comes back.

What still might get labeled

Images made with Google’s Imagen carry SynthID, and Meta AI’s own images carry Meta’s watermark. Those can be detected without any metadata. Meta also runs classifiers on uploads. A clean file removes the most common trigger, the metadata, and that is what fixes false labels on real photos edited in Photoshop or Lightroom. It does not turn a generated image into a photograph, and Meta’s rules still require disclosure of realistic synthetic content.

Stories, Reels and carousels

The same detection runs on every surface. Stories and carousel slides are labeled per image, so clean every file in the set. For Reels and other video, Instagram relies on disclosure and its own detection; this tool handles still images only.

Real photo, wrong label?

If Photoshop’s Generative Fill, Lightroom’s AI features or your phone’s Magic Editor touched a real photo, the file now says AI was involved. Read why real photos get AI labels and strip the metadata before posting.

Questions people ask

Can I remove the AI label after posting?

Not directly. Delete the post, clean the image with the tool above and upload the clean copy. Reposting the same file will trigger the same label.

Does Instagram strip metadata anyway?

Yes, but only after reading it. The label decision happens during upload, so metadata that is present at that moment counts, even though the published image no longer contains it.

Will screenshotting the image work instead?

A screenshot has no metadata and different pixels, so it avoids the metadata trigger, but it also loses resolution and color accuracy. Re-encoding with a 3% shrink gives the same effect at full quality.

Is removing the label against Instagram’s rules?

Meta requires disclosure of realistic AI-generated or altered content. Correcting a false label on a real photograph or removing private metadata is fine. Using a clean file to pass off synthetic media as real breaks the rules.