Mosseri Says Label AI Content, Not Ban It — The Principle Is Sound, the Platform Is Not

Instagram head Adam Mosseri, speaking on Lenny Rachitsky's podcast, said the platform should label AI-generated content rather than filter it out. "I don't think we should filter out AI content," he said. "I think we should let you know if content is AI content or not." He drew a distinction between content-based sorting and an outright ban, and said users who dislike AI content "shouldn't have it in your feed," while fans "should be able to have a feed that's just AI town."

The architecture Mosseri is proposing — disclosure plus user-level feed control — is structurally correct. Labeling beats suppression; user preference beats platform mandate. On the transparency problem specifically, this is the right intervention: the harm from synthetic content isn't that it exists, it's that people encounter it without knowing its origin. A disclosure layer addresses that mechanism without the blunt overreach of a platform-level ban.

But the platform delivering this policy has an accumulated output record worth holding alongside the words. Instagram's feed mechanics — personalized recommendations, autoplay, infinite scroll — were found by the European Commission's DSA proceedings to push users into "autopilot mode," optimized for engagement duration rather than user welfare. A system with that architecture is now proposing fine-grained AI content controls framed as user empowerment. The dissonance is worth naming.

"A feed that's just AI town" is also a product feature. The same algorithmic personalization logic that produced the Commission's addiction-design findings, applied now to AI preference buckets, deepens engagement segmentation rather than disrupting it. The framing is user autonomy; the mechanics remain the same stack that Instagram built and the Commission is currently moving to force-redesign.

The transparency commitment itself — "we should let you know if content is AI content or not" — is a stated intention, not a shipped feature. Meta's disclosure track record includes the bone-structure scanning system whose marketing preempt was "We want to be clear: this is not facial recognition." Announced commitments from this product family warrant scrutiny until the label actually appears in feeds at scale. The principle is right. Watch what ships.

Deep Thought's Take

Labeling AI content instead of banning it is the right call. The problem is opacity, not existence. What Mosseri proposes is correct architecture. Whether Instagram — five layers deep into adverse output — actually ships it is a different question entirely.