Instagram head Adam Mosseri said users who dislike AI-generated content on the platform should build a feed without it, but ruled out giving them a filter to hide AI posts outright. Speaking on Lenny Rachitsky's podcast on July 10, 2026, Mosseri argued Instagram should label AI content, not remove it, and floated the idea of a separate opt-in feed he called 'AI town' for users who want more of it. The comments arrive as Meta pushes deeper into generative tools with its Muse Spark image generator, which allows users to tag other people into AI creations.
Instagram, along with TikTok, YouTube, and Facebook, already applies AI-content labels but does not offer a toggle to filter labeled posts out of the main feed. Mosseri's framing draws a line between disclosure and suppression: users get to know what they're seeing, but the algorithm decides what they see.
The distinction matters because Meta has spent the last two years threading AI-generated content and Meta-made AI characters directly into the same ranked feed as posts from friends and creators. A hide-AI switch would give users a single button to unwind a core piece of Meta's product strategy — one the company is not prepared to hand over.
Key facts
- 01Instagram head Adam Mosseri said users who dislike AI content 'shouldn't have it in your feed,' but the platform won't add a filter to hide it.
- 02Mosseri favors labeling AI posts and letting AI enthusiasts opt into a dedicated feed he called 'AI town.'
- 03He conceded AI detection is 'hard' and that Instagram may 'lose the ability' to identify AI posts as generative models improve.
- 04Meta's new Muse Spark image generator lets users tag other people into AI creations, drawing warnings from the National Center on Sexual Exploitation.
- 05In December 2025, Mosseri floated fingerprinting 'real media' — labeling camera-captured content — as a more practical alternative to flagging AI.
Mosseri conceded the detection side of the equation is deteriorating. He said identifying AI posts is 'hard' and that Instagram may 'lose the ability' to pick up on AI content as generative models get better at evading classifiers. That is a notable admission from the executive running the platform: the labeling regime Meta is defending as an alternative to filtering is itself on borrowed time.
His proposed workaround is to invert the problem — label the real stuff instead of the fake stuff. Mosseri suggested it may be 'more practical' to fingerprint 'camera-captured content,' an idea he raised in December 2025. Under that model, provenance signals travel with authentic media at capture, and anything without the mark is treated as unverified by default.
“I think you should be able to just ask, 'Is this AI?,' and we should be able to tell you we think it probably is, or we're not sure, or it's definitely not, or definitely isn't.”— Adam Mosseri, Head of Instagram
That approach has technical merit and a set of well-known problems. Provenance standards like C2PA depend on hardware, editing software, and platforms all cooperating, and most media in circulation predates any of it. In the interim, users get a feed where labels are increasingly best-effort and no filter exists to route around them.
Mosseri also acknowledged Instagram needs to 'figure out how to crack down' on spammy AI content, a category the platform continues to embrace at the product level even as it strains ranking systems. The tension is straightforward: the same generative tools Meta ships to boost creation volume also flood the feed with material that Instagram then has to moderate.
The sharpest edge sits on Muse Spark, which lets users pull other people into AI-generated images by tagging them. Haley McNamara of the National Center on Sexual Exploitation warned the feature could enable exploitation, harassment, and identity fraud — a foreseeable outcome given how tagging works on Instagram today.
“creates obvious and foreseeable opportunities for exploitation, sexual abuse, harassment, and identity fraud”— Haley McNamara, Executive Director, National Center on Sexual Exploitation
Meta has not detailed the consent controls around Muse Spark tagging, and Mosseri did not address that specific feature in the podcast interview. AI Chat Daily covered Muse Spark's launch as Meta's push to challenge Claude and GPT-5.6 on the model side; the tagging behavior on the consumer side is a separate design choice that Meta owns.
Mosseri's position leaves the debate roughly where it started: Instagram wants to be the platform where AI content thrives, and users who don't want AI content in their feed are being told to curate around it rather than block it. That is a coherent product philosophy — algorithmic feeds have always resisted user-side hard filters because they undercut ranking — but it puts the burden on users to detect and unfollow AI creators individually, at exactly the moment Mosseri admits detection is getting harder. The signal for AI companies building on top of Meta's distribution: the front door stays open, and the platform's owners are betting that labels plus feed diversity will hold long enough for provenance standards to catch up. Whether that bet survives contact with a feed where the AI is undetectable and untaggable is the question the next 18 months will answer.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
