New York accuses TikTok of serving users ‘placebo’ safety features

Algo Refresh is presented as a way for TikTok users to reset their recommendation algorithm. That matters because the algorithm influences which videos appear in a person’s feed. A reset could help users move away from unwanted or harmful recommendations.

The feature was described as an algorithm-control tool. New York alleges that some users believed they had activated it, but TikTok gave them a nonworking version instead. Their feeds did not change, unlike feeds belonging to users who received the fully working feature.

The article also says TikTok allegedly presented Algo Refresh as a permanent reset, even though the reset was temporary. That could affect how users understand the control they are being offered. The feature’s purpose is therefore both practical and transparent: it should change recommendations and accurately explain what that change means.

New York alleges that TikTok gave selected users nonworking versions of safety features. These versions were described as “placebo” or “ghost” features because they appeared available but did not provide the promised protection or control. The allegation is part of an updated version of New York’s lawsuit.

Algo Refresh is the clearest example. Some users allegedly believed they had reset their algorithms, but their feeds did not change. Other users received the fully working version, and their feeds did change. New York also alleges TikTok withheld other safety improvements from selected users during similar tests.

TikTok spokesperson Nathaniel Brown said the company regularly tests products and features to understand how they work in the real world. The dispute centers on whether testing a feature is acceptable when users are not clearly told that they may receive a version that does not work.

New York alleges that thousands of TikTok users were affected by the tests. The group reportedly included children and teenagers, making the allegations especially significant because young people were among those who may have believed they were using working safety controls.

The article does not give an exact total. It describes the number only as “thousands,” and it does not break that figure down by age, country, or test. The allegations concern users who were allegedly given nonworking versions of safety features, including Algo Refresh.

The scale matters because the issue may not have involved only a small technical trial. If the allegations are proven, many users may have believed they had more control over their feeds than they actually did. New York’s claims are contained in an updated, sealed version of its 2024 lawsuit, later ordered unsealed by a judge.

The alleged test created different outcomes for different users. People who received the nonworking version believed they had activated Algo Refresh, yet their TikTok feeds stayed the same. That meant the feature appeared to offer control without producing its expected result.

Users who received the fully working version had a different experience. Their feeds changed after they used the feature. The contrast between the two groups is central to New York’s allegation that TikTok provided a “placebo” rather than a real safety feature to selected users.

The article does not describe exactly how the feeds changed or which videos appeared afterward. It says only that the feeds of the placebo-group users did not change, unlike those of users who received the working version. This alleged difference is why the test raises questions about transparency and meaningful user control.

New York argues that calling Algo Refresh a permanent reset could mislead users about what the feature actually does. A permanent reset suggests that the user’s changed recommendation settings will continue indefinitely. The article says the reset was temporary instead.

That difference matters because users may decide whether to trust or use the feature based on its expected duration. Someone seeking lasting control over their feed could believe the problem had been permanently addressed, even though the effect would not continue that way. The article does not specify how long the temporary reset lasted.

The allegation adds a second concern beyond whether the feature worked. New York claims TikTok both tested nonworking versions on some users and described the feature inaccurately. TikTok says it regularly tests features, but the lawsuit challenges whether users received clear and truthful information about Algo Refresh.

An A/B test on a digital platform compares different versions of a feature with different user groups. A placebo group receives a version designed to look similar but does not include the active change. Companies can use these tests to measure how features work in real-world conditions.

The controversy is sharper when the feature concerns safety, privacy, or user control. People may rely on the feature’s apparent protection or assume their choices have taken effect. In this case, New York alleges that users believed Algo Refresh had reset their feeds even when it had not. A TikTok product manager reportedly objected that a placebo group conflicted with the feature’s purpose of transparency and control.

TikTok spokesperson Nathaniel Brown defended regular product testing in general. The lawsuit raises the narrower question of whether users should receive an inactive safety feature without clear disclosure, especially when children and teenagers are among the affected users.

Recommendation algorithms are systems that select and rank content for each user. In general, they use signals such as viewing behavior, likes, shares, follows, searches, and similar interactions to estimate which videos a person may want to watch. TikTok then presents a personalized feed based on those predictions.

The article does not list TikTok’s exact signals or explain its ranking formula. It does establish that an algorithm shapes users’ feeds and that Algo Refresh is intended to reset it. New York alleges that some users believed they had reset their algorithms, but their feeds did not change.

That alleged failure matters because recommendation systems can influence whether users encounter unwanted or harmful content. A working reset could potentially help users move away from such recommendations. This explanation of algorithm mechanics uses established general knowledge; the article itself only describes the algorithm’s role and the alleged effect of Algo Refresh.