You posted a photo and Instagram labeled it AI Generated. What happened?

Instagram scans each image at upload time for specific metadata signals. The primary one is a C2PA content credentials manifest — a cryptographically signed block of data embedded in the file that identifies which AI tool created or modified it. If that manifest includes a trainedAlgorithmicMedia digital source type, Instagram applies the "AI Info" label automatically. The label was previously called "Made with AI" before August 2024.

This process is fully automated. No human reviews the image before the label appears. Instagram cannot tell whether you intended the label or whether it is a false positive from a minor AI-assisted edit.

PrivyClean showing C2PA content credentials detected in an image file

What triggers the AI Info label

Several types of metadata can cause Instagram to apply the label:

  • C2PA content credentials from tools like ChatGPT/DALL-E, Adobe Firefly, Google Gemini, and Microsoft Copilot/Designer. These are cryptographically signed manifests stored in JUMBF blocks inside the file.
  • IPTC Digital Source Type set to trainedAlgorithmicMedia — the standard tag indicating AI generation, used by Midjourney, many stock photo platforms, and other tools.
  • XMP tags identifying AI generation tools — the xmp:CreatorTool field may read "Adobe Firefly" or "DALL-E," which platforms also scan.
  • Adobe Creative Cloud apps — Photoshop and Lightroom embed C2PA content credentials whenever any AI feature is used, including Generative Fill, Generative Expand, Neural Filters, AI-powered Denoise, and AI Masking.

That last point is the source of most false positives. A real photograph edited with a single Generative Fill stroke receives the same C2PA tag as a fully AI-generated DALL-E image.

The false positive problem for photographers

Instagram has no way to distinguish between "this image was 100% generated by AI" and "this is a real photograph where the photographer removed a power line with Generative Fill." Both produce the same C2PA tag. Both get labeled.

This is a serious issue for working photographers, wedding shooters, and commercial photographers who use AI-powered editing tools as part of their normal workflow. A client receiving an edited photo portfolio may see AI labels on real photographs they commissioned for authentic documentary purposes.

Once the label is applied to a published post, it cannot be removed from that post. The only option is to delete the post and re-upload with a clean file.

How to prevent the AI Info label before posting

  1. Inspect the file before uploading. Open the image in PrivyClean to check whether C2PA or AI-related metadata is present. The app groups metadata by category so you can quickly see if AI content credentials are attached.
  2. Decide whether to clean. If the AI metadata is a false positive from a minor edit, you may want to remove it. If the image is genuinely AI-generated and you want transparency with your audience, keep it.
  3. Export a clean copy. PrivyClean exports a new file with the selected metadata removed. Your original file stays untouched.
  4. Upload the clean copy to Instagram. The cleaned file no longer carries the metadata signal that triggers the label.

If you export from Photoshop or Lightroom, you can also disable Content Credentials in the export settings before generating the file. But this only works if you remember before exporting — PrivyClean handles the case where you already have the exported file and want to inspect it.

For a full explanation of C2PA and which metadata fields are involved, see our plain-English guide to C2PA content credentials.

When you should keep the AI Info label

There are situations where keeping AI metadata is the right call:

  • The image is fully AI-generated and you value transparency with your audience about how it was made.
  • You are distributing content in the EU and want to comply proactively with AI disclosure requirements ahead of the August 2, 2026 EU AI Act deadline.
  • Your editorial or journalistic standards require disclosure of AI involvement, even minor.
  • Your audience or client has specifically asked for provenance information to be retained in the file.

Metadata control is not the same as deception. The goal is to make an intentional, informed choice — not to remove metadata blindly. Learn more about how this affects different types of creators in our EU AI Act overview.

What about taking a screenshot instead?

Screenshots strip metadata but also significantly reduce image quality. Instagram compresses images at upload; starting from a screenshot adds another compression step before that. The result is noticeably lower quality compared to uploading the original file with metadata cleaned properly.

Re-saving in some apps (like Apple Photos) may strip some metadata, but the behavior is inconsistent across versions and file types. A dedicated metadata tool gives you visibility into what was actually removed before you upload.

Related guides

Important: Users are responsible for complying with applicable laws and platform terms of service. Removing AI content credentials from images that were substantially AI-generated may conflict with platform disclosure policies or future regulatory requirements.