Instagram is slapping its "AI Content" label on ordinary, human-edited photos, while letting fully AI-generated images pass through untagged. The mislabeling matters because it undermines the one signal platforms offer to help people tell real content from synthetic content.
What actually happened
Users have reported Instagram auto-applying the "AI Content" label to photos edited only with Canva's Background Remover or minor blemish tools, according to The Verge. One Threads user said the tag appeared "every time there is bg remover involved." Canva told content strategist Jess Bruno that some of its assistive tools "were being tagged as generative" and said this is now fixed, though some users still report mistagging afterward. Cosmetics brand About Face, founded by singer Halsey, had recent photos tagged despite its social manager stating: "This photo was taken on my iPhone and then slightly edited by myself in the photo app." In reporter Jess Weatherbed's own tests, images fully generated in Google Gemini, carrying both C2PA and SynthID watermarks, were not tagged at all. Only content made with Meta's own AI app triggered the label. A test Instagram account posting AI images in quick succession went untagged for nearly two weeks, even after Meta announced a crackdown on AI accounts. Meta did not respond to requests for comment.
How we got here
This is not Instagram's first labeling failure. A similar problem hit its "Made by AI" tag in 2024, months after launch. In February 2024, Meta said it would scan images for IPTC and C2PA metadata to detect generative AI use. That system reportedly flagged photos with only minor Adobe-tool retouching, prompting Meta to promise adjustments reflecting "the amount of AI used in an image." Meta has also said it checks for "industry standard indicators" from other companies' tools, without detailing what those are or how scanning works. Two years later, the underlying opacity remains, and the errors have resurfaced in a new form.
Why this matters for you
For everyday users, the label's unreliability means neither a tag nor its absence can currently be trusted as proof of how an image was made. For creators and brands, mistaken tags risk reputational harm, as seen with About Face publicly denying AI use. For builders working on provenance tools, this is a live case study in why metadata standards like C2PA and SynthID need consistent platform-side enforcement, not selective scanning. As AR and wearable devices increasingly capture and share images directly, reliable content-origin labeling becomes more, not less, important for the ecosystem.
The bigger question
If a platform cannot reliably tell real photos from AI-generated ones, and admits only vague scanning methods, what should users, or regulators, treat as a trustworthy signal of authenticity going forward?
What to watch
No fix date has been announced by Meta. Canva says its tagging issue is resolved, though user reports of continued mistagging persist. The Verge's test account and its untagged AI images remained live and public as of the report. Whether Meta issues a public explanation of its detection criteria remains an open item to track.
