TikTok Ran 'Ghost' Safety Experiments on Teens, New York Alleges in Unsealed Lawsuit

THURSDAY, OCTOBER 8, 2026 — Imagine this: you're a teenager drowning in a TikTok feed that's showing you things you don't want to see, so you reach for the safety lever the company gave you — "Algo Refresh," the tool designed to reset your recommendations and let you escape. You pull it. Your feed looks the same. You assume that's just how it works. According to allegations unsealed Wednesday in New York Attorney General Letitia James's lawsuit against TikTok, that's exactly what the company was counting on: for thousands of users, including teens and children, the lever wasn't connected to anything.

The allegations, reported by Reuters, describe so-called "ghost" or placebo experiments. TikTok selected thousands of users who had opted to use safety features like Algo Refresh and served them non-functional versions of the tool. Those users believed they had activated the feature, but their feeds did not change — unlike users who got the fully working version. The experiments, according to the unsealed complaint, were run repeatedly in 2023, with the company measuring how long people stayed on the platform and the effect on advertising revenue.

What the unsealed allegations say

The details in the unsealed complaint are granular, and they are ugly. According to the Britain Herald's account of the filings, the "ghost" experiments were not a one-off test: they were run repeatedly across 2023, each time on thousands of users. The metrics being measured were engagement and ad revenue — how long people remained on the platform — even as the test subjects believed they had just activated a safety feature meant to give them more control.

Perhaps the most striking allegation is that someone inside the company flagged it. An internal TikTok product manager for Digital Well Being reportedly raised concerns about the placebo tests, arguing that giving users a non-functional safety feature conflicted with the purpose of a tool designed to provide greater control over recommendations. The concern was raised, the complaint alleges — and the tests continued.

The complaint also alleges that TikTok misrepresented how Algo Refresh worked in the first place, claiming the feature would reset a user's recommendation algorithm when the effect was allegedly temporary. And there is a second, darker set of claims: separate experiments in which safety protections limiting exposure to harmful content were switched off for randomly selected users. The unsealed allegations reference a New York teenager who was allegedly exposed to large amounts of content related to suicide and self-harm while participating in one of the tests.

TikTok's response, and the states' case

TikTok has broadly denied New York's allegations in court filings and has said that keeping young users safe is a priority. In a statement, a company spokesperson said: "We continuously make improvements to our platform which means that, like most companies, we regularly test new products and features to validate the experience and understand how they work in the real world."

The lawsuit itself is one of more than two dozen brought by states accusing TikTok of designing its platform to addict children and misleading users about its safety. Many documents in state cases across the country have been sealed or heavily redacted as TikTok has sought to keep details about its operations from public view — which is why the unsealing of these allegations lands with such force. The suit claims TikTok publicly assures users that safety is a top priority and its platform is appropriate for kids and teens, while knowingly preventing some groups of users from accessing certain safety tools — including users who had specifically sought to avoid unwanted or potentially harmful content.

This didn't start Wednesday

The ghost-experiment allegations have been building for weeks. In August, Bloomberg Businessweek reported the existence of the experiment from a confidential 2023 company document — and Bloomberg Law reported that Senators Marsha Blackburn and Richard Blumenthal sent a letter to TikTok CEO Shou Chew demanding answers, calling the decision to conduct the test "depraved." Blackburn and Blumenthal are co-sponsors of an online child safety bill, and their letter framed the experiment as one that withheld a safety feature from millions of users, including a teenager who died by suicide.

Then, in September, TikTok settled the first state lawsuit set for trial — agreeing to pay Alabama up to $300 million and change how its platform functions for teens in the state. The New York allegations, unsealed October 7, now give the public the most detailed look yet at what exactly the state attorneys general believe TikTok did behind closed doors.

Why this one lands differently

Tech companies A/B test everything; nobody disputes that. What's different about the ghost-experiment allegations is the asymmetry. A/B testing a button color affects everyone equally. Running a placebo test on a safety feature means the control group isn't a control group — it's people in distress who asked for help and were given the appearance of it. The lawsuit's framing is that these users did exactly what the company asked them to do — use the safety tools — and the company used the moment to measure ad revenue.

The case is still allegations, and TikTok's denials are on the record. But the unsealed documents have now put a term into the public conversation — "ghost experiments" — that is going to be very difficult to unsee, in courtrooms or anywhere else.

More culture coverage: October's TikTok Halloween costume trends, the Buldak noodle world record, and Nicole Kidman at Chanel's Paris Fashion Week show.