NY AG: TikTok Accused of Secret 'Algo Refresh' Teen Test

The New York Attorney General has accused TikTok of secretly testing a product feature labeled "Algo Refresh" on teenage users without disclosure, according to a report carried by Escudo Digital.

What does "Algo Refresh" actually mean?

The Escudo Digital headline identifies the feature as an algorithmic test — a controlled variation in how the platform's recommendation engine (the system that ranks and serves videos in each user's feed) ranks or selects content for teen accounts. The limited reporting available does not spell out the mechanics: whether the feature resets user preference signals, whether it offers any opt-out (a user-facing control to decline participation), or how long it ran. Treat the name as a placeholder until TikTok or the Attorney General's office publishes the technical detail.

What's confirmed vs. what's alleged

The accusation itself is on the public record, attributed to the New York Attorney General's office. The technical scope, the test window, and the size of the affected teen cohort remain undisclosed. No direct statement from TikTok responding to the allegation appears in the available Escudo Digital reporting. Practitioners reading the headline should hold two ideas at once: a regulator has named the company, and the underlying product behavior has not yet been independently documented.

Why this lands now

State attorneys general have spent recent years pressing social platforms on teen-facing defaults — from audio-on settings to screen-time dashboards (in-app panels that summarize minutes spent). A regulator now naming a specific, allegedly undisclosed algorithm test sharpens the precedent: experimentation on minors requires disclosure, not just good intent. The accusation puts TikTok's internal research pipeline squarely in view.

What does this change for marketers?

For brands running teen-adjacent campaigns — verticals (product categories) whose audience overlaps with under-18 viewers — the practical question is reach integrity. If TikTok's recommendation engine shifted mid-flight during the alleged test, attribution models (the frameworks that credit an ad view with a downstream action like a click or purchase) built on pre-test performance will misfire. Buyers should quarantine recent TikTok engagement data for under-18-leaning segments until the test parameters come into view.

CPM (cost per thousand impressions, the price an advertiser pays per 1,000 ad views) benchmarks for those segments may also need a recalibration once the test window is known. Even a brief algorithmic swing can pull average view-through rates (the algorithm of users who watched an ad past a defined duration) by enough points to skew planning models.

What does this change for creators?

Creators who depend on teen viewership for breakout velocity (the algorithm speed at which a new video reaches non-followers) should ask: did the test suppress my reach to under-18 viewers during a defined window? If so, the algorithm may have buried videos that would have performed differently outside the test. There is no creator-side tool today that would tell a publisher whether they were inside or outside the test cohort.

Until TikTok publishes a response or the Attorney General's office files a formal action, the safest move is documentation — screenshotting performance anomalies (unusual spikes or drops) and flagging them internally.

The road ahead

Watch for two signals in the coming weeks: TikTok's first public statement on "Algo Refresh," and any filing from the New York Attorney General's office that names the test parameters. The first will tell marketers whether the company treats the feature as a bug or as a defensible product choice. The second will tell agencies whether to revise teen-segment planning assumptions. Both matter. Neither has appeared yet.