The EU accuses TikTok's 'rabbit hole' algorithm of breaching its content rules

TikTok’s recommendation algorithm is the system that selects personalized videos for each user’s For You page. It uses signals from a user’s activity to predict which videos might keep that person engaged. This matters because the feed can strongly shape what users repeatedly see.

For example, if someone watches and interacts with videos about body image, TikTok may recommend more videos on that subject. The article says the more a user interacts with a certain kind of content, the more similar videos they see. This repeated pattern can create a “rabbit hole,” especially when the topic involves sadness, isolation, or body image.

The European Commission says TikTok should rely more on explicit preferences and stop using viewing time as a recommendation indicator. It also wants more accessible controls and regular algorithm testing. TikTok rejects the findings and says users have direct control over their content experience.

The European Commission accuses TikTok of breaching the Digital Services Act because its For You recommendation system may expose users to harmful content repeatedly. The regulator says TikTok did not do enough to assess how its algorithm could pull people into “rabbit holes” or protect their wellbeing. Vulnerable adults and minors may face particular risks.

The Commission points to the system’s personalized design. When users interact with a type of video, they may receive more similar videos. It also says tapping “not interested” may not be enough to change the feed. In the regulator’s view, TikTok did not put sufficient measures in place to reduce these risks.

The findings are preliminary, so the EU has not yet made a final determination. TikTok disputes the claims, calling the Commission’s statements inaccurate. If the findings are confirmed and TikTok does not resolve the issue, ByteDance could face a major fine.

A harmful “rabbit hole” can form when an algorithm repeatedly reinforces the same kind of content. Each interaction gives the system another reason to recommend related videos. The concern is that users may become surrounded by one emotional or sensitive subject instead of seeing a wider range of content.

For instance, a person who watches videos about sadness may receive more videos about sadness, isolation, or related experiences. The article describes similar risks for body-image content. Because the feed is personalized, repeated exposure can become endless, and the Commission says this may have a detrimental impact on mental health, especially for vulnerable adults and minors.

The EU says TikTok did not sufficiently assess or reduce this danger. It also says the “not interested” button may not adequately change recommendations. The Commission wants stronger user controls, less reliance on viewing time, and regular testing of the algorithm.

If the European Union confirms the Commission’s findings, ByteDance could face a fine of as much as six percent of its annual global revenue. The possible penalty relates to an alleged breach of the Digital Services Act involving TikTok’s recommendation system. Its scale makes the dispute significant for TikTok’s parent company.

The fine is not automatic. The Commission has issued preliminary findings, meaning the matter has not reached a final conclusion in the article. TikTok would have an opportunity to address the concerns. The potential penalty becomes relevant if the EU decides the violations are confirmed and the company has not sufficiently resolved them.

The Commission is asking TikTok to redesign parts of the For You page and strengthen safeguards. Requested changes include using explicit preferences more, removing viewing time as a recommendation signal, improving controls, and testing the algorithm regularly. TikTok rejects the regulator’s claims and defends its safety measures.

The European Commission wants TikTok to redesign how the For You page recommends videos. Its goal is to reduce the chance that users are drawn into harmful, repeated content patterns. The regulator says recommendations should reflect users’ explicit choices more clearly, rather than relying so heavily on behavior that may be difficult to control.

Specifically, the Commission wants TikTok to rely more on explicit preferences and disable viewing time as an indicator for recommending content. It also wants user controls to become more accessible and easier to use. The article notes that pressing “not interested” may currently be insufficient to change a person’s feed.

Finally, the Commission wants TikTok to test its algorithm regularly. These requests form part of preliminary findings in a wider investigation. TikTok says the Commission’s statements distort its platform architecture and safety work. If the EU confirms the findings, the company could face a fine of up to six percent of global annual revenue.

The EU and other regulators are focusing on children and teenagers because the article links social media features to mental-health, privacy, and addiction concerns. The Commission warns that recommendation “rabbit holes” may especially harm minors and vulnerable adults. Regulators around the world have therefore placed social media companies under closer scrutiny.

The article gives several examples involving TikTok. In February, the EU said the app breached the DSA because of its “addictive design.” In July, officials said TikTok failed to protect the privacy of minors’ accounts. The latest preliminary findings concern possible harms from the For You recommendation system.

The scrutiny is broader than the EU. Google, Meta, and TikTok sued California over a law banning platforms from letting users under 16 access potentially addictive features. These developments suggest regulators will continue examining how platform design affects younger users and what safeguards companies provide.

Recommendation algorithms learn from a user’s behavior by treating actions as clues about interests. Interacting with a video can signal that its topic matters to the user. In TikTok’s case, the article says more interaction with one kind of content leads to more similar videos appearing on the For You page.

Viewing time can be a powerful signal because it measures sustained attention. A video watched for longer may be interpreted as more engaging than one quickly skipped. If the system uses that signal, it may recommend more content with similar themes. The resulting feedback loop can keep narrowing the feed around one topic.

The Commission specifically wants TikTok to disable viewing time as a recommendation indicator. It argues that this could help reduce harmful rabbit holes. The article does not describe TikTok’s full technical training process, but it reports that TikTok disputes the Commission’s account of its platform architecture.