TikTok’s AI cleanup is moving beyond labels
An account-detection test targets spam on sensitive subjects, while a separate control adjusts the amount of AI content people see.
TikTok is testing a way to spot accounts built around AI-generated spam, extending its response beyond labels on individual videos. In a September 29 Irish newsroom update, the company described improved detection aimed at subjects including politics, current events, financial advice and medical content—areas it links to risks to public trust or well-being.
The same announcement described tests of Manage Topics controls that let people adjust how much AI-generated content they see. That is a preference about the feed, separate from detecting spam accounts. TikTok did not describe an off switch that removes all AI content, or provide results establishing how effective the account test has been.
The company’s October 28 APAC update adds scale to its disclosure effort: TikTok said it had labelled more than three billion videos as AI-generated through content credentials, creator tools and invisible watermarking. Those labels identify material created or significantly altered with AI. They are not a count of spam removals or evidence that every labelled video violated a rule.
TikTok also said automated systems accounted for 97.4% of all video removals globally in the second quarter of 2026. That broader moderation figure does not measure the September AI-spam test. For that test, availability, effectiveness and resulting removals remain unestablished in the cited announcements.