Why Instagram Shows “AI Info” on Your Real Photo — and How to Prevent It

You took the photo yourself. You only removed a stray power line or brightened the sky. Yet when you post it, Instagram adds “AI info” — and the comments start asking whether it’s fake.

This guide explains what Meta’s label means, what actually triggers it, why real photos get caught, and how to check your files before you post.

What the “AI info” label is

Meta — which runs Instagram, Facebook and Threads — labels images and videos that it believes were created or edited with generative AI.

• In early 2024 Meta began applying a “Made with AI” label.
• Photographers quickly complained that real photos were being labelled after ordinary edits. In mid-2024 Meta renamed the label “AI info”, saying the old wording didn’t make clear that an image might only have been edited with AI tools rather than created by them.
• In September 2024 Meta changed where the label appears. For content that was only edited with AI tools, the information moved into the ⋯ menu on the post. Content Meta believes was generated by AI keeps a more visible “AI info” label.

Tapping the label shows a short explanation and, where available, a way to ask for a review.

What triggers the label

Meta has said it relies on industry-standard signals embedded in the file, plus self-disclosure and its own detection. In practice, the label is triggered by one or more of the following:

1. C2PA Content Credentials

C2PA is a signed record, stored inside the image, that describes how it was made. Many AI image generators add it automatically, and so do some editing apps when you use a generative feature. If the manifest says an AI tool created or modified the image, the platform can read that and apply a label.

2. IPTC “Digital Source Type”

The IPTC photo metadata standard includes a field called Digital Source Type. Values such as trainedAlgorithmicMedia (“created by AI”) or compositeWithTrainedAlgorithmicMedia (“a composite that includes AI elements”) tell platforms that AI was involved. Editing software writes this into the file’s XMP metadata.

3. Self-disclosure

Meta asks creators to disclose realistic AI-generated video and audio, and you can add an AI label yourself when posting.

4. Detection

Some generators also embed invisible watermarks in the pixels (Google’s SynthID is one example), and platforms run their own classifiers. These signals are part of the image itself, not its metadata.

Why real photos get labelled

The most common reason is simple: your editing app told Instagram it used AI. Features like generative fill, generative remove, AI expand, sky replacement or object removal may be powered by a generative model. When you use them, the app can record that fact in C2PA and IPTC/XMP metadata — even if the change was tiny.

The label then reflects the tool you used, not how much of the image was changed. A 99%-real photo can carry exactly the same signal as a fully generated one.

How to check a photo before posting

1. Export the photo from your editing app as you normally would.
2. Open it in the metadata viewer. It runs in your browser — nothing is uploaded.
3. Look at the red AI footprints box at the top. It names the signals found, such as C2PA, IPTC Digital Source Type, or specific tools like Firefly or Google AI.
4. Expand the tag groups to see exactly which fields mention AI.

If the box is green (“No AI-generator tags found”), the file itself carries no AI metadata. If it’s red, you know what the platform is likely to see.

How to avoid the label on genuine photos

If your photo is real and you only used AI-assisted tools for minor clean-up, you have three honest options:

1. Use non-generative tools for small fixes. Classic healing, clone and spot-removal brushes don’t usually create AI provenance records.
2. Check your app’s export settings. Some editors let you choose whether to include Content Credentials when exporting.
3. Remove the metadata before posting. The TraceWipe metadata remover strips C2PA, IPTC, XMP and EXIF without re-compressing your image, then re-scans the result so you can confirm nothing is left. As a side benefit, it also removes GPS location and device details.

What removing metadata does — and doesn’t — do

Be clear about the limits:

• It removes metadata signals (C2PA, IPTC, XMP, EXIF).
• It does not remove pixel watermarks such as SynthID, and it doesn’t change what the image looks like.
• It does not change platform detection. Platforms may still label content based on their own analysis.
• It does not change a post that’s already live. To remove a label from an existing post, delete and re-upload the cleaned file, or use the platform’s review option.

Use it responsibly

Metadata exists partly so people can understand where content comes from. Removing a misleading tag from a genuine photo is reasonable; hiding that an image is substantially AI-generated is a different matter. Meta’s rules ask creators to disclose realistic AI-generated content, and regulations — including the EU AI Act’s transparency obligations — increasingly require it. If an image is mainly AI-generated, label it.

Frequently asked questions

Can I remove the “AI info” label after posting?

Not directly — there’s no switch to turn it off. You can request a review if you believe it was applied by mistake, or delete the post and upload a version without the AI metadata.

Will taking a screenshot remove the label?

A screenshot is a new file without the original metadata, so it often avoids metadata-based labels. But it lowers quality and doesn’t affect pixel watermarks. Removing the metadata from the original file keeps full quality.

Does Lightroom or Photoshop add AI metadata?

Using Adobe’s generative features can add Content Credentials and related XMP data. Check your export settings and inspect the exported file to be sure.

Does this apply to Facebook and Threads too?

Yes. The same labelling system is used across Instagram, Facebook and Threads.