How Does TikTok's Algorithm Work? What Attorneys Need to Know About the For You Page

TikTok's algorithm works by ranking videos for each viewer individually, using an interest graph built from watch behavior rather than a social graph built from follower relationships. The For You page predicts what a specific user will watch next based on signals like watch time, completion, rewatches, likes, shares, and follows. For attorneys, that one design choice changes how reach, virality, and audience exposure have to be analyzed in litigation.

This is the part of TikTok litigation where I spend most of my expert time, because it is the part that platform outsiders most often get wrong. With more than 200 million American users receiving individually ranked feeds, there is no single "audience" that saw a video, and follower counts tell you almost nothing about who actually did.

What Signals Drive the For You Page?

TikTok has published descriptions of how its recommendation system ranks content, and the structure has been consistent even as the models evolve. The system weighs three families of signals:

• User interactions: videos watched to completion, rewatched, liked, shared, or commented on, accounts followed, and content created. TikTok's published explanations indicate that strong interest signals, such as finishing a longer video, outweigh weak ones, such as the viewer and creator being in the same country.

• Video information: captions, sounds, hashtags, and effects, which help the system classify what a video is about and which interest clusters it might serve.

• Device and account settings: language, country, and device type, which TikTok describes as lower-weight inputs used mainly for basic relevance.

Two consequences matter for litigation. First, distribution is earned per video, not per account. A creator with 200 followers can reach millions if early viewers watch to the end, while an account with a large following can post a video that goes almost nowhere.

Second, feeds are personal. What the algorithm served to one user is evidence about that user's interaction history, not just about the video.

Why Does the Interest Graph Matter in Court?

Because most litigation intuitions about audience come from the social graph era. On a follower-based platform, reach roughly tracks audience size, so follower counts and impressions feel like a measure of exposure. On TikTok, the interest graph breaks that assumption:

• View counts are not audience identity. A million views tells you how many times the video was served and watched past the threshold TikTok counts, not who saw it, where, or in what context.

• Virality is not intent by itself. Whether a poster meant to reach a mass audience is a separate question from whether the algorithm delivered one. I analyze both in my work, including forming opinions on what a poster's platform behavior indicates about intended audience.

• Reach is reconstructable. Analytics available to the account holder, records obtainable through legal process, and the video's engagement trajectory allow an expert to build a defensible account of how content actually spread.

In my expert witness work on TikTok matters, reach analysis is where cases are won and lost. A defamation plaintiff claiming reputational harm to a specific community, an employer arguing a workplace video was seen by customers, or a prosecutor asserting a threat reached its target all need more than a view counter. They need an explanation of distribution, and that explanation has to match how the For You page actually behaves.

How Has the 2026 U.S. Transition Changed the Algorithm Question?

On January 23, 2026, TikTok's U.S. operations moved to TikTok USDS Joint Venture LLC, a majority American-owned entity in which ByteDance retains 19.9 percent. Under the joint venture's announced structure, the U.S. recommendation system is operated under U.S. oversight, retrained on U.S. user data, and run in Oracle's secure U.S. cloud environment, with the joint venture holding decision-making authority over trust and safety policies.

For litigators, that transition has practical consequences. Questions about how the U.S. algorithm behaved after the transition are now questions about the U.S. entity's system, which affects discovery targets, the entity to which process is directed, and how expert opinions about pre-transition and post-transition distribution should be framed. Where a case spans the transition, the distinction belongs in the analysis, and I flag it in my own reports.

Where Does Algorithm Analysis Show Up in Litigation?

• Product liability and platform accountability cases, where what the recommendation system served to a user, and how repeatedly, is a central factual question.

• Defamation, where damages turn on actual reach into the relevant community rather than raw view counts, and where the resurfacing behavior of recommendation systems can extend the life of a statement.

• Criminal matters, where the defense question is often whether a defendant sought out content or was served it, a distinction the interest graph makes analytically meaningful.

• Influencer and contract disputes, where guaranteed reach, engagement authenticity, and underperformance claims require separating algorithmic variance from breach. Account valuation raises the same issues, which I cover in what an Instagram account is worth in a dispute.

• Employment and harassment matters, where whether coworkers or customers plausibly saw a video depends on distribution analysis, not assumption.

The common thread is that the algorithm is not a black box excuse in either direction. Its published mechanics, its observable outputs, and the records that legal process can reach support real analysis. Courts do not need to take anyone's word for how content spread. That analysis pairs with the foundational evidence questions I cover in whether TikTok videos can be used as evidence in court and how TikTok evidence is used in court.

What Should Attorneys Ask For in Discovery?

When distribution matters, the records worth pursuing early include the account's own analytics, the video-level engagement trajectory over time, audience breakdowns available to the account holder, and, through appropriate process, platform records bearing on distribution. Pair those with preservation, since analytics windows and retention limits can trim what is available. An expert who has worked with these records, and I do this work on active TikTok engagements, can tell you within a short consultation whether the reach story a case depends on is provable, overstated, or simply wrong.

Frequently Asked Questions

How does TikTok's For You page decide what to show?

TikTok's published explanations describe ranking based on user interactions such as watch time, completion, likes, shares, and follows, combined with video information like sounds and captions, and lower-weight device and account settings. Feeds are ranked per viewer, not broadcast to followers.

Does follower count determine TikTok reach?

No. TikTok distributes each video on its own performance through the interest graph. Small accounts can reach millions and large accounts can reach almost no one, which is why follower-based reach assumptions fail in litigation.

Can you prove who saw a TikTok video?

Individual-viewer identification is generally not available from public data, but reach can be reconstructed in meaningful ways using account analytics, engagement records, distribution patterns, and records obtained through legal process, which supports expert opinions about audience and exposure.

Who controls TikTok's algorithm in the United States?

Since January 23, 2026, TikTok's U.S. operations, including operation of the U.S. recommendation system, sit with TikTok USDS Joint Venture LLC, a majority American-owned entity, with U.S. user data held in Oracle's U.S. cloud environment under the venture's announced structure.

About Kate Talbot, TikTok Expert Witness

Kate Talbot is a social media expert witness retained in 15+ cases across 10+ law firms, spanning intellectual property, employment, personal injury, defamation, and federal litigation. Her platform work covers Instagram, TikTok, Snapchat, YouTube, and X, including active engagements on TikTok platform mechanics.

If TikTok reach, virality, or algorithm behavior matters to your case, review her social media expert witness services or request a consultation.