EU says TikTok's For You feed breaches the Digital Services Act over 'rabbit hole' effect

The European Commission has told TikTok that its main "For You" video feed appears to breach the EU's Digital Services Act (DSA), because the way the recommendation algorithm works can draw users into so-called "rabbit holes" of similar content about sadness, isolation or body image. The preliminary findings, announced in Brussels on Friday 6 November 2026, say TikTok did not properly assess that risk and that the safeguards it introduced have not worked well enough.

The Commission says the feed's design needs to change. TikTok, which is owned by China's ByteDance, has rejected the findings. Nothing is final yet: the company can now reply, and only a later non-compliance decision could lead to a fine, which under the DSA can reach up to 6% of a company's total worldwide annual turnover.

What did the European Commission find?

The Commission's press release (IP/26/2361) describes the mechanism in plain terms: "The more a user interacts with a certain type of content, the more TikTok's 'For You' feed algorithm recommends similar content." It says this "can be harmful when such endless and repeated exposure involves, for example, content about sadness, isolation or body image," and that the resulting rabbit holes can harm mental health, "especially that of vulnerable adults and minors."

The findings come in two parts, matching two core duties that the DSA places on the largest platforms.

On risk assessment, the Commission preliminarily found that TikTok "did not adequately assess" how its recommendation algorithm may lead users into rabbit holes. It explains that the algorithm uses engagement signals such as watching and liking videos to personalise recommendations. If a user repeatedly watches or likes content that may be triggering, such as videos about weight loss or body image, the system reads that as interest and recommends more of the same. According to the Commission, repeated exposure of this kind may contribute to sadness or anxiety, body dissatisfaction, lower self-esteem, or the worsening of existing eating disorders and other mental health conditions.

On risk mitigation, the Commission acknowledged that TikTok has already put some measures in place but found them "not effective". Features meant to diversify the feed do not sufficiently reduce how often similar content appears, it said, and the control tools do not let users effectively shape what they are shown. Its example is the "not interested" button, which "may not be enough to change the content feed" because the algorithm prioritises viewing time over a user's stated choice to see less of something.

The Commission said its view rests on an in-depth investigation, including TikTok's own risk assessment reports, internal data and documents, the company's answers to multiple information requests, a review of scientific research, and interviews with experts in recommender systems, psychology and psychiatry. Ireland's media regulator, Coimisiún na Meán, supported the investigation; TikTok's European headquarters is in Dublin, RTÉ noted.

What does Brussels want TikTok to change?

At this stage, the Commission "considers that TikTok needs to change the design of the 'For You' feed." It listed examples of what that could involve:

• relying more on users' explicit preferences rather than inferred interest;
• disabling viewing time as a signal for recommending content;
• making user controls easier to find and use;
• regularly testing the algorithm.

The second point is notable. Viewing time, how long someone lingers on a video, is one of the signals short-video apps commonly use to decide what to show next. A policy analysis published by Tech Policy Observer on 8 November pointed out that this goes further than general calls to "adapt the design" of a service, because it targets a specific ranking parameter. That analysis also noted that the preliminary findings themselves are not public; outside observers are working from the press release and the text of the regulation.

Henna Virkkunen, the Commission's Executive Vice-President for Tech Sovereignty, Security and Democracy, said: "Every time a user scrolls or likes a video on TikTok, the algorithm serves more of the same leading users down rabbit holes they cannot easily escape from." She added that the DSA "makes platforms responsible for addressing the negative effects related to their recommender system."

How has TikTok responded?

TikTok disputes the Commission's account. "The Commission has made a series of statements that are fundamentally inaccurate and distort our platform's architecture and our commitment to user safety," a spokesperson said in a statement reported by Euronews and RTÉ. "We have a strong record of protecting our community while giving users direct and intuitive control over their content experience."

The company has pushed back on earlier findings in the same case too. Reuters reported that when the Commission charged TikTok in February over what it called addictive design, TikTok described those findings as "categorically false" and said it would challenge them.

What is the Digital Services Act and why does it apply here?

The Digital Services Act (Regulation (EU) 2022/2065) is the EU's rulebook for online platforms. Services designated as "very large online platforms", those with more than 45 million monthly users in the EU, carry the heaviest obligations, and TikTok is one of them.

Two articles are central to this case. Article 34 requires very large platforms to identify, analyse and assess "systemic risks" stemming from the design or functioning of their services, explicitly including algorithmic systems. Those risks include "serious negative consequences to the person's physical and mental well-being" and effects on the protection of minors, and the assessment must consider how the design of recommender systems influences them. Article 35 then requires "reasonable, proportionate and effective" mitigation measures, which may include adapting a service's design and testing and adapting its recommender systems.