Why does a real photo get an AI label?
Platforms read tags inside the file at upload. Instagram and Facebook show “AI content” (formerly “AI info”), TikTok shows “AI-generated”, Pinterest shows “AI modified”, and LinkedIn shows a Content Credentials “CR” icon. None of them measures how much of the picture changed. They see that an app wrote an AI tag.
Editing apps write these tags even for small AI edits. A wedding photo with one stranger removed, or a product shot with a cleaner background, can carry the same kind of tag as an image generated from scratch. The label describes the file, not your camera.
This affects people who never think of themselves as AI users: a parent who removed a stranger with the phone’s eraser, a photographer who ran AI Denoise on a night shot, a shop owner who cleaned up a product background.
The scan above tells you more than “AI found”. It shows what your file claims. In IPTC terms, compositeWithTrainedAlgorithmicMedia means a real photo combined with AI edits, while trainedAlgorithmicMedia means an image created by AI. If yours says edited and the edit was small, the label says more than what really happened.
Which editing tools leave an AI tag in a photo?
Phone edits are easy to forget, because the eraser in the gallery app looks like an ordinary editing tool. These apps are common sources of AI tags in photos:
- Photoshop and Lightroom: Generative Fill, Generative Remove and AI Denoise
- Adobe Firefly and Adobe Express
- Canva Magic tools
- Google Photos Magic Editor and Magic Eraser, including on Pixel phones
- Samsung Galaxy AI edits
- Apple Clean Up and Image Playground
- Images made or edited in ChatGPT, Microsoft Designer or Copilot, and Midjourney
How to remove AI tags from a photo in 3 steps
In the default mode the tool removes AI tags only and keeps camera details, location and the color profile. Turn on the privacy option to also remove location and camera details. The steps:
- Drop the edited export, not the untouched original. The AI tag is written when the app saves the edit, so the export is the file that carries it.
- Check whether the file says “edited with AI” or “created by AI”, and which app wrote it. That tells you whether the label is false or simply accurate.
- Tap “Delete AI label” and post the clean copy. On a phone, share it straight to the app from the result screen instead of picking the original from your gallery.
Do I need to convert HEIC, PNG or WebP first?
No. Photos straight from an iPhone in HEIC can be cleaned as they are, and PNG and WebP files keep their format too. Converting a photo to another format, or taking a screenshot of it, compresses the image again and can lower its quality. Here the file keeps its format and the pixels stay identical, because only metadata blocks are cut out.
Some images also store the prompt and settings used to make them in text fields inside the file. Those are removed along with the other AI tags, which also keeps that information private.
Already posted? What to do about a label on a live post
No file tool can change a post that is already published. On Instagram and Facebook there is no switch to remove an automatic label. Request a review where it is offered, or delete the post, clean the original file and post again, which loses the likes and comments. On Pinterest, use the appeal process for labels you believe are wrong.
If you added the label yourself with the AI toggle when posting on Instagram or Facebook, edit the post and turn it off. Always repost from your original file, not from a copy saved from the app.
Before you repost, drop the original above to confirm the tag was the cause. If the scan finds no AI tags, the label came from somewhere else, and posting the same file again will not change it.
How to stop the AI tag at the source
A few habits keep the next photo clean from the start:
- Photoshop and Lightroom: Adobe says that since August 2026 its apps attach Content Credentials automatically after generative AI edits, so the export setting may not keep them out. Check each export here.
- Save a copy before the AI step, so you can choose which version to post.
- Google tools, such as Reimagine in Magic Editor and Gemini, add an invisible SynthID watermark with no switch to turn it off.
- Samsung’s generative edits can also place a small visible sparkle mark in a corner of the image. That mark is part of the picture, and this tool does not change pixels.
SynthID, and what cleaning leaves to the platform
Google’s SynthID watermark sits in the pixels, and cleaning removes it: if you edited your own photo with a Google AI tool, one click takes out both the AI tags and the SynthID mark, so the clean copy carries no AI marks. You do not need any detector to check first.
Two things still stay as they are after cleaning, because both live outside the file in a way we cannot reach:
- Visible marks, such as the sparkle that some phone editors add to generative edits.
- The verdict of AI detector websites, which judge the pixels and can disagree with each other about the same photo.
When should a photo keep its AI label?
If a picture is fully AI-generated, keep the disclosure. This tool is for real photos that were tagged because of an editing step, and for keeping your location and device details private. Three questions help you decide:
- Was the image generated from a prompt rather than taken with a camera? Keep the label.
- Does it show real people or events in a way that did not happen? Keep the label or disclose it yourself.
- Is it a real photo with a small AI fix, such as a removed object or less noise? The label overstates the edit, and cleaning the file is reasonable.