Meta reads industry metadata to label AI content. Here is what triggers the tag and what to do when a real photo gets flagged.
🔒 Your files never leave your device — processing happens in your browser.
How Meta decides
Meta labels content on Instagram, Facebook and Threads in three main ways: when you choose to disclose AI yourself, when its systems detect industry-standard signals in the file — C2PA Content Credentials and IPTC metadata declaring AI generation — and when its own detection flags the content. Images made with Meta AI are labelled automatically.
Why real photos get labelled
Editing apps add AI signals even for small edits. Using generative fill to remove a cable, AI noise reduction, an AI upscaler or an AI sky replacement can write a C2PA manifest or an IPTC “composite with trained algorithmic media” tag into a photo that is essentially real. Meta then reads the tag on upload.
What you can do
- Check the file with the checker to see which signal is present.
- If the image is a genuine photo with a minor retouch, clean the metadata with the tool above before posting.
- If the image is substantially AI-generated or photorealistic AI, keep the label — Meta asks users to disclose realistic AI content and can apply penalties if you don’t.
Note that removing metadata cannot affect Meta’s own visual detection or invisible watermarks.
Frequently asked questions
Can I remove the AI info label after posting?
On many accounts you can edit the post’s AI label setting, but labels applied from detected signals may stay. Cleaning the file before upload is what changes the metadata Meta reads.
Does a screenshot remove the label?
A screenshot creates a new file without the original metadata, but it lowers quality. Clean Image Metadata removes the metadata while keeping the original quality.