Instagram's AI Labels Are Flagging Real Photos and Missing Actual AI Images

Key points

• Instagram's "AI Content" label has been incorrectly applied to real photographs edited with basic tools like Canva's Background Remover, affecting ordinary users since at least early 2025.
• Canva confirmed its assistive editing tools were being tagged as generative AI in error, and says it has fixed the problem on its end, though some users still report false flags after the fix.
• About Face, a cosmetics brand founded by singer Halsey, had iPhone photos flagged as AI-generated despite the brand's social manager saying no AI tools were used.
• Meta has not explained which signals it scans for, and did not respond to requests for clarification from The Verge.
• A journalist's own test account, posting images made with genuinely AI-generated content, was never flagged at all.

If you've posted a photo to Instagram recently and seen a small "AI Content" label appear on it, you're not alone, and you may not have done anything that warrants it.

Over the past several weeks, users have reported that Meta is stamping ordinary edited photos with that label, treating them as synthetically generated content: images built from scratch by artificial intelligence software rather than captured by a camera. At the same time, pictures that actually were made by AI tools are getting through without any label.

Why is this happening?

Nobody knows for certain, and Meta isn't saying.

The label is supposed to work by reading invisible digital certificates embedded in an image file, certificates that record whether a generative AI tool touched the image. Two of the main certificate systems are IPTC and C2PA, industry-standard formats for logging what happened to a file. The trouble is that plenty of everyday editing tools also write to those certificates, even when they do nothing remotely generative. Canva's Background Remover uses machine learning, a form of AI that has been inside editing apps for over a decade, to cut out a background cleanly. That's a far cry from asking software to conjure an image of a beach that never existed.

Content strategist Jess Bruno investigated the Canva connection and found that some of Canva's assistive tools were writing metadata that Instagram read as proof of generative AI use. Canva told her the problem was fixed, and a note now appears on Canva's background-removal help page saying the tool doesn't add AI-generated content metadata to a file. Yet some users report false labels persisting even after Canva said it had corrected the issue.

We first covered the C2PA certificate standard on 4 September 2026; it's now at the centre of a labelling dispute that affects anyone who edits a photo before posting.

What does this mean for ordinary Instagram users?

For most people, a mislabelled photo is embarrassing rather than harmful. For businesses or creators whose credibility rests on showing real people and real work, a false AI label carries a real cost.

About Face had recent product photos labelled as AI-generated on Instagram. The brand's social manager replied publicly that the images were shot on an iPhone and "slightly edited" in the standard Photos app. It's not clear which iPhone editing tools were used, though only a small number of Apple's newer AI features, including Spatial Reframing, Extend, and an updated Clean Up introduced in iOS 27, add generative AI certificates to a file.

A journalist from The Verge ran controlled tests, posting images edited with Canva, Photoshop, Adobe's Firefly generative AI features, and Google's Gemini AI tools. None were labelled. The only images that triggered the label were ones made with Meta's own AI app.

That's a striking gap. The system appears most reliable at detecting content made with Meta's own tools and least reliable at everything else. Our earlier story from 31 August 2026 reported Instagram's plan to penalise accounts that hide AI-generated personas: a worthwhile goal, but one that depends on a detection system that currently can't tell Canva's eraser tool from a text-to-image generator.

What should you do if your photo gets wrongly labelled?

The clearest practical step is to check which editing tools you use before posting. Background-removal and object-selection tools may write AI-related metadata even when the result looks identical to the original photo.

If your image is flagged incorrectly, Instagram does allow users to report labelling errors, though Meta has offered no timeline for fixing the broader detection problem.

The deeper issue is transparency. Meta has never published a clear list of the signals it scans for, citing the risk of bad actors gaming the system. That argument has merit. But right now the people being tripped up aren't bad actors. They're a shop owner posting a product shot, a nurse sharing a work anniversary photo, a small brand trying to show its work is real. A label that can't be trusted either way helps no one.

Common questions

Does using Canva automatically make Instagram flag my photo as AI?

Not always, but it has happened to many users, particularly with Canva's Background Remover tool. Canva says it has corrected the metadata issue on its end, but some users report false flags persisting.

Can I stop my photos from being mislabelled?

There's no guaranteed fix right now. Avoiding tools that use any form of automated editing lowers the risk, but Meta hasn't published a definitive list of what triggers the label, so it's difficult to know exactly what to avoid.