TikTok adds AI smart keyword filters to clean up feeds

TikTok just gave users a smarter way to clean up their For You feed.

The company announced a new AI-powered feature called Smart Keyword Filters, designed to block unwanted content more effectively by automatically catching synonyms and related terms.

Outsmarting feed fatigue

Standard keyword blockers have always had a glaring loophole: block one word, and creators simply swap in a synonym to bypass it. TikTok says users have already filtered over 200 million words globally, but the cat-and-mouse game never stopped.

The new Smart Keyword Filters use AI to automatically identify and restrict content containing words related to the ones you enter. The interface shows you exactly which additional terms the AI decided to filter out.

Over the coming months, TikTok plans to let users manually check or uncheck those AI-suggested terms to fine-tune the filter. The platform is also doubling the personal keyword limit to 200 words and introducing bulk upload tools.

Alongside the keyword upgrades, TikTok rolled out a "Manage Topics" dashboard. It lets users adjust how often they see clips from more than 10 broad categories, including Creative Arts, Travel, Nature, and Sports, dialing recommendation frequency up or down over time.

Both teen and adult accounts will get access to the features, which TikTok is pairing with a new educational guide and explainer videos created with safety advocate @cathypedrayes.

Why it matters

If your TikTok feed feels stale or hijacked by repetitive trends, this update gives you a much faster way to reset your boundaries without having to train the algorithm from scratch with constant "not interested" taps.

Strategically, this is a clever move for TikTok. Algorithmic feeds have faced intense scrutiny from regulators and mental health advocates for trapping users in addictive or harmful rabbit holes. By introducing AI-expanded mute lists and category sliders, TikTok hands the illusion and reality of control back to the user without breaking its core recommendation engine.

The risk, however, lies in AI overreach. Synonyms are rarely one-to-one matches, and an aggressive model could easily filter out videos you actually wanted to see. The fact that TikTok is already promising a way to deselect specific AI-added terms shows they know false positives will be a real problem.