Why Your Real Photo Got an “AI Info” Label

Real photos sometimes get labelled as AI after small AI-assisted edits. Here's which metadata causes it, how to check your file, and what your options are.

You take a real photo, remove a stray object with a generative fill tool, post it, and it appears with an "AI info" label. It's a common frustration for photographers, and it usually comes down to metadata, not the pixels.

Why platforms add the label

Several large platforms label content they believe is AI-generated or AI-edited. Meta's apps, for example, renamed their label from "Made with AI" to "AI info" in 2024, after photographers complained that lightly edited photos were being tagged.

To decide when to show a label, platforms combine several signals. One of the main ones is industry-standard metadata in the file:

  1. C2PA Content Credentials. A signed manifest that can record AI actions, such as generative fill or expand, in an image's history. See our guide to Content Credentials.
  2. The IPTC "digital source type" field. Stored in XMP, it can say the image was trainedAlgorithmicMedia (fully AI-generated) or compositeWithTrainedAlgorithmicMedia (a real photo combined with AI-generated elements).

Editing apps write these fields automatically when you use certain AI-powered tools, even if the change was tiny.

Edits that can add AI metadata

  • Generative fill, expand or remove in photo editors
  • AI background replacement or sky replacement
  • Some AI-powered denoise or upscaling tools, depending on the app and version
  • Exports from apps and templates that use generative features behind the scenes

Not every AI-assisted feature writes provenance metadata, and behaviour changes between app versions. The only reliable way to know is to check the exported file.

How to check your photo

Drop the exported file into the MetaZero cleaner. Each image gets an AI label line telling you whether it contains:

  • a C2PA manifest, and which software signed it
  • an IPTC digital source type that indicates AI involvement
  • generator names in fields like Creator tool or Software

If none of these show up and the image still gets labelled, the platform is probably using other signals, such as its own image analysis or an invisible watermark embedded in the pixels. Metadata tools can't change those.

Your options

Disclose it. If AI meaningfully changed the image, a label may simply be accurate, and many platforms require you to disclose realistic AI-generated or AI-altered content. Check the rules of the platform you're posting to.

Avoid adding it in the first place. For edits you could do with traditional tools such as cloning, healing or cropping, using those instead won't add AI provenance metadata.

Clean your own file. If the label misrepresents your photo, you can remove the provenance metadata from your own image before posting. In MetaZero:

  1. Choose Lossless strip to keep every pixel identical.
  2. Tick Keep camera details if you want to keep make, model, lens and exposure information. C2PA, XMP and IPTC data are removed. GPS and serial numbers are removed too.
  3. Download the file and check the result line says it's clean.

Keep your original. Always keep an unmodified master copy. If you ever need to prove how a photo was made, the original with its full edit history is your best evidence.

The bottom line

"AI info" labels on real photos are usually triggered by metadata that editing tools add automatically. Check what's in your file before you post, decide whether the label is accurate, and follow the platform's disclosure rules.