TikTok's ‘For You’ Feed Is Under Fresh EU Scrutiny. What Does That Mean for Your Feed?
On 6 November, European regulators raised fresh concerns about how personalised video recommendations can repeat sensitive themes. The finding is preliminary, and TikTok disputes it. Here is what is known—and what users can control now.

AI-generated editorial illustration, reused from Safegram News Article 62; a conversation about online choices, not a scene from the EU investigation.
Written by Safegram Editorial Team
Key takeaways
- On 6 November 2026, the European Commission issued preliminary, not final, findings about the risk of harmful content “rabbit holes” in TikTok's For You feed. [1]
- The concern is the feedback loop that can repeatedly recommend content on sadness, isolation or body image to people who engage with similar videos; the Commission has not shown that every viewer is harmed. [1]
- TikTok disputes the criticism and offers controls such as Not interested, keyword filters, feed refresh and topic management. Those options are not evidence that the wider systemic risk has been resolved. [3][4]
What changed on 6 November?
The European Commission said it had preliminarily found a breach of the EU Digital Services Act (DSA) over the harmful “rabbit hole” effect of TikTok's personalised For You video feed. The notice focuses on whether the company sufficiently assessed and reduced the risks of repeatedly recommending similar sensitive content. The Commission cited mental-health struggles and body-image themes, particularly for minors and vulnerable adults. [1]
This is not a final infringement decision, a court judgment, a blanket ban on recommendations or a finding that every TikTok session damages wellbeing. TikTok can contest the allegations as the regulatory process continues. TikTok told Euronews that the Commission's descriptions were fundamentally inaccurate and pointed to its existing protections and user controls. [4]
What is a recommendation “rabbit hole”?
A recommendation feed chooses what to show next using signals including viewing and interaction patterns. A rabbit hole describes a feedback loop: watch or engage with a certain theme and the system offers more of it, which creates further engagement signals. This can help people discover artists, skills and communities. The concern becomes sharper when a loop repeatedly amplifies potentially distressing themes without enough variety or friction. [1][3]
A hypothetical example: someone searches for a video about body confidence and is then shown one similar clip after another. This does not mean the person is ill or that every clip is harmful. It illustrates the mechanism regulators want platforms to evaluate, especially for people who may be vulnerable. No individual case is asserted here.

AI-generated editorial illustration, reused from Article 62: creators may find audience discovery useful while also wanting greater control.
Why the DSA matters, and what it does not say
Under the DSA, very large online platforms must assess and mitigate systemic risks associated with the design and use of their services. The Commission's 6 November communication concerns this regulatory duty—not a new EU law banning TikTok or requiring every feed to be chronological. In February 2026, the Commission separately issued preliminary findings about potentially addictive design features, including infinite scroll, autoplay and notifications. Those earlier concerns remain distinct from the November focus on recommendation loops. [1][2]
The public notice does not publish a precise, independently verified rate of harm, an Ireland-only user count or a guaranteed remedy. A risk assessment is about patterns and safeguards, not proof that each viewing event causes a particular outcome. It is also not evidence that rival platforms are immune: personalisation, useful discovery and meaningful control can coexist—or conflict—across the sector.
What TikTok users can change today
TikTok's own help materials describe several controls, though labels and availability may vary by account, region and app version. Start with Not interested on individual videos; explore Filter video keywords and Manage topics; and consider Refresh your For You feed if recommendations no longer match your interests. In supported EU experiences, non-personalised feed options may also be available. Use TikTok's current Help Centre rather than assuming this list is universal. [3]
None of those buttons guarantees that distressing content disappears. They are settings you can test, not a substitute for the platform's obligation to evaluate systemic risks. For families, discuss what a young person is actually seeing without assuming a technical control replaces conversation and age-appropriate support. If content is causing distress, taking a break or seeking appropriate support matters more than optimising an algorithm.

AI-generated editorial illustration, reused from Article 62: asking what a recommendation system is optimising for.
The question for creators: discovery without a trap
Recommendation systems also introduce small businesses, independent artists and educators to people who would not otherwise find them. That is a genuine benefit. Yet creators may feel pressure to maximise watch time or repeat the same emotional format, while audiences may struggle to leave an intense stream. A fairer design question is whether users can easily choose a different feed and understand why content appears.
For a fictional Dublin fitness instructor, a short video being recommended to new viewers might create useful discovery. The same instructor should not have to promise extreme outcomes or exploit anxiety to find an audience. This is a hypothetical illustration, not a claim about TikTok's ranking criteria or anyone's earnings. Trustworthy social platforms should give meaningful choice and readable safety controls, without pretending that verification or any single setting solves every problem.

AI-generated editorial illustration, reused from Article 62: individuals and businesses have a stake in safer, more explainable feeds.
Does this affect Ireland, Britain, Brazil or China equally?
The DSA is EU law, so the new Commission action is an EU enforcement matter, relevant to services operating in the Union, including Ireland. The UK has a separate online-safety regime; readers in Brazil, the US or mainland China should not assume this specific EU regulatory process grants them the same remedies or says anything about local availability. Platforms can offer some similar settings internationally, but the exact interface should be checked locally.
Frequently asked questions
Has the EU banned TikTok's For You algorithm?
No. On 6 November 2026 the Commission announced preliminary findings about a possible DSA breach. This is not a final decision or a ban on recommendation feeds. [1]
How do I reset or influence TikTok's recommendations?
Check TikTok's Help Centre for Refresh your For You feed, Not interested, keyword filters and topic-management controls. Availability and effectiveness vary; no setting guarantees a particular result. [3]
Does watching similar videos prove an addiction or mental-health harm?
No. Repetition can be a warning sign worth assessing, but the Commission's statement does not diagnose individuals or establish that every viewer is harmed. If viewing leaves you distressed, pause and seek support appropriate to your circumstances.
Sources and further reading
- European Commission — Preliminary findings on TikTok's For You “rabbit holes,” 6 November 2026
- European Commission — Preliminary findings on addictive design, 6 February 2026
- TikTok Help Centre — How TikTok recommends content
- Euronews — TikTok disputes the Commission's 6 November findings
What readers can do next
Test the feed controls your platform actually provides, notice which themes keep repeating, and ask for transparent explanations when a recommendation system shapes your attention. Read more independent coverage of privacy, safety and digital trust at Safegram News.
More from Safegram
Try Safegram
Privacy-first social and a verified marketplace, built in Dublin.