Snap announced on 31 July that wholly AI-generated videos are no longer eligible for recommendation on Spotlight, its short-form vertical feed.
The clips can still be posted. They simply will not be shown to anyone who does not already follow the account, which on a discovery feed amounts to the same thing.
The stated reasoning is unusually direct for a platform policy document: as low-quality, repetitive, AI-generated content spreads across the internet, Snap wants Spotlight to remain somewhere people find original perspectives and personal storytelling.
Cutting to the chase, this is a commercial judgment dressed as a values statement. And it is the right one.
Let me explain. Snapchat still reports 971 million monthly users and retains an unusual hold on younger audiences, the group most likely to walk away from a feed that feels machine-assembled.
The company also says unique Spotlight contributors are up more than 120% year on year, which gives it something rarer than a policy: an actual supply of humans to protect.
Compare this with Meta, which launched an entirely AI-generated video app in February and labels synthetic posts rather than demoting them. And it's had its share of controversy.
It would seem the two companies have placed opposing bets and only one of them has to be right. But have they really?
Snap's position is smarter than a blanket prohibition would be, because the carve-out does the work.
Videos enhanced or edited with Snapchat's own AI tools stay eligible, with transparency indicators attached.
The message to creators is that AI can help you make something but cannot be the whole thing, which happens to sit neatly with keeping people inside Snap's editing suite.
The obvious objection is enforcement, and Snap concedes it, acknowledging that no detection system is perfect.
But there's a catch. Detecting fully generated video at scale is not currently possible, the word "wholly" is doing enormous load-bearing work, and a great deal of AI can go into a clip before it crosses that line.
Meaningful enforcement requires standardised provenance metadata, something like Google's SynthID, adopted by camera manufacturers, editing software and every major platform simultaneously.
The industry has understood this for years and shipped nothing universal, partly because the companies best placed to build it are the ones selling the generators.
Snap does not have the leverage to force a standard. Its nuclear option, refusing any upload without an official watermark, would be commercially absurd for a company its size.
None of which makes the announcement empty. It establishes that a public company now believes human-made content is worth protecting as an asset rather than tolerating as a cost, which is a reversal of the assumption that has governed feeds for a decade.
The EU's AI Act transparency rules took effect on 2 August, so regulation is arriving regardless.
Snap has simply worked out that being early looks like principle rather than compliance.
