TikTok Not Interested button: what the audit measured

What did the audit actually measure?

The paper is dated 11 May 2026. It studies the For You Page on accounts the researchers controlled. It says the most effective explicit signal — marking a video as Not Interested — is unintuitively buried in the interface. It also says that once accounts cease to indicate disinterest in a topic, many find their feeds dominated by such content again.

Northeastern Global News (byline June 9, 2026) describes the same audit. The researchers used emulation software and had each account watch more than 200 videos related to a specific topic via search and their For You page. One topic was cooking, one was fitness, and one was sports betting.

Ars Technica (Jul 8, 2026) says their methodology did not involve computer simulations; they created bot accounts on the actual TikTok mobile app. Kaplan told Ars they used emulated devices.

Is Not Interested stronger than swiping, and by how much?

On the paper, the cooking cell is this paragraph. On-topic videos constituted 16% of the implicit signaling account’s feed, compared 30.5% of the feed of the account which continued watching them (a 47.5% decrease). Explicit signaling led to an even more pronounced decrease at 4.75% (a total decrease of 84.4%). The next paragraph says that was on one topic (cooking), then including all three target topics and running Phases 2 and 3 five times per topic.

Ars Technica prints a shorter pair: the button reduced unwanted content by around 84 percent, compared to just a 48 percent reduction from merely skipping videos. Around 84 percent is not the paper’s 84.4%. A 48 percent reduction is not the paper’s 47.5% decrease.

Northeastern Global News prints a run count. In three out of the five tests on both cooking and fitness, as part of Phase 2, marking a video as Not Interested was more effective than swiping. For sports betting, that was the case in four out of the five runs, Kaplan said. The paper’s line is that whether explicit signaling is more effective than implicit signaling differs across topics and runs.

Kaplan, on the news page: the Not Interested button is better than just swiping past content, and users should use it, but if they find that their feed is still dominated by content [they don’t like], our research shows that this is something that can happen.

Where is the button, and how long does the tap last?

The paper says the button is accessible rather nonintuitively via the ‘Share’ menu or by long-pressing a video, and may not be known to all users. It calls that signal unintuitively buried in the interface. Northeastern Global News says the platform website describes the button as something you bring up after long pressing on a post. Ars Technica says the authors note that the option seems to be deliberately hidden from users.

What should a creator do with a signal that can relapse?

The paper: We then use these techniques to study the FYP algorithm based on accounts we control. Kaplan, on the news page, says users should use it, and that a feed can still be dominated by content [they don’t like].

The news page says topics may begin popping up again within minutes after stopping. The paper says many accounts which cease to express disinterest and begin watching topical videos again can see their feeds dominated by such videos. That behavior is more common for cooking and fitness than for sports betting. Ars says even a brief re-engagement by a user is sufficient, and that if you start watching it, the platform will again feed it to you more.

The audit’s topics were cooking, fitness, and sports betting. A separate page covers organic TikTok growth without ads. Another covers the follower-first ranking change. Nearby is a different feed surface, covered in the local discovery rollout.

1. 1

   Use the control the paper names

   The paper says ‘Share’ menu or long-pressing a video. The news page says long pressing on a post.
2. 2

   Expect the news clock and the paper condition

   Northeastern Global News says users may need to continually click, because topics may begin popping up again within minutes after stopping. The paper says accounts which cease to express disinterest and begin watching topical videos again can see their feeds dominated by such videos.
3. 3

   Read the topic split before you generalize

   Cooking and fitness relapsed in the explicit case twice on the paper. Sports betting was found to relapse less often. The news page says cooking in particular appeared quickly after the accounts stopped pressing the button.

The paper also says a different sock-puppet account ran on a different device. Devices in each column are run simultaneously.