Instagram head Adam Mosseri says anger about AI will keep growing

Adam Mosseri went on the Colin and Samir podcast to talk about long-form video and spent part of it on how much people hate what his industry is shipping. The hosts asked about college students booing tech executives at graduation ceremonies. Mosseri said a lot of people dislike AI, and not only young people and creatives, and that he sees more anxiety, more frustration and more anger about it, strongest in creative work.

“It's not only getting worse, it's going to continue to get worse”

— Adam Mosseri, head of Instagram, on the Colin and Samir podcast, 28 October 2026, https://www.theverge.com/ai-artificial-intelligence/1021644/adam-mosseri-instagram-ai-negativity

On ranking he gave two answers that do not fit together. Instagram does not rank AI accounts differently from anyone else. Then he asked himself whether AI content ranks as well as the rest and answered probably not. He also said there is plenty of good AI content alongside the bad, and that people throw the word slop around too liberally.

Two admissions, one pressure

A year ago at Bloomberg's Screentime conference Mosseri said generative AI would reduce the cost of producing content to basically zero. This week he said the money Instagram pays video creators runs ROI negative: the company spends more than it makes back by a long shot, and scaling the program up would be burning capital.

Put the two together. When production costs fall to nothing, a bonus paid per view hands the budget to whoever can generate the most clips, and a feed that ranks AI content like everything else fills with it. Mosseri said paid creators do not post much more, that the extra they post is often weaker, and that payments end up moving views from one video to another rather than adding any. When he answers probably not, Instagram is leaving the demotion to the audience instead of writing a rule.

Labels, deepfakes and one of his own carousels

Instagram added a label for AI accounts earlier this year and began removing AI-generated profiles in August. It now requires AI characters to label themselves, and Mosseri said banning AI content outright is not practical. Automatic detection runs inference over every upload and misfires: one of his own carousels came back classified as AI content. The complaint that fills his direct messages is stolen stand-up material performed by an AI character, and he said originality is the principled answer, which would let Instagram treat a copied bit as unoriginal whoever delivers it.

Long-form video was the reason for the interview. Instagram announced a test of uploads up to 90 minutes on 26 October and an expansion to Apple TV in the US. Mosseri wants pre-roll and mid-roll ads, because they make payouts easy to attribute to a specific video, and he expects to take heat for them from viewers who find them more annoying than ads between short clips. He said Instagram is late to the party and will keep cutting bespoke deals to bootstrap the ecosystem.

Numbers exist for the anger he describes. Researchers at Edinburgh and Bath counted 90,835 machine-written responses out of 118,756 in a UK vape consultation, and the government published an outcome that counted only the other 27,921. Vitalik Buterin said this week that a model detects any reference and then reads a text at the surface. Both are the complaint Mosseri hears from creators: the volume arrives, the judgment does not.

Informational material. Quotes come from the Colin and Samir interview as reported by The Verge and NetInfluencer; the 2025 cost-of-content line comes from TechCrunch's report on Bloomberg Screentime. Reading the two admissions as answers to the same pressure is ours.