How the TikTok Algorithm Works

What TikTok has said publicly

TikTok has described its For You feed as personalised for each user, based on a combination of factors. In its public explanations it has grouped these broadly into:

• User interactions: videos you like, share, comment on, watch fully or skip, and accounts you follow.
• Video information: details like captions, sounds and hashtags.
• Device and account settings: such as language preference and country, which TikTok has described as carrying less weight.

Why watch behaviour seems so important

Because the feed is built around individual viewers, how people react to a video — watching it, rewatching, sharing or swiping away — is a natural input. That's why advice about hooks and retention keeps coming up: it's the part of the system creators can influence most directly.

Shares and comments as a strong signal (observation)

Many creators report that videos which get shared to friends or spark discussion in comments tend to keep getting pushed to new viewers for longer than videos with similar view counts but little interaction. TikTok hasn't published an exact weighting for this, but it fits with the platform's stated interest in interactions beyond simple watch-through.

Follower count matters less than you might expect

TikTok has said that follower count and previous high-performing videos aren't direct factors in recommendations, though accounts with more followers naturally get more views from those followers. This is why smaller creators can sometimes reach large audiences with a single video.

Diversity and repetition

TikTok has also said it tries to avoid showing too much of the same thing and may interrupt repetitive patterns. That helps explain why a creator's reach can vary a lot from one video to the next.

Why one video can hugely outperform the next

Creators often notice that a single video from an otherwise quiet account suddenly reaches far more people than anything else they've posted. This is consistent with a system that tests each video with a small initial group of viewers and expands distribution based on how they respond, rather than relying heavily on the account's history. TikTok hasn't detailed the exact mechanics of this testing process, so treat it as an informed observation rather than a confirmed rule.

What creators can control

Common misconceptions worth retiring

• "TikTok has a shadowban feature that silently limits accounts." TikTok doesn't describe its systems this way; ineligibility for recommendation is typically tied to specific guideline violations.
• "Deleting and reposting a flopped video resets the algorithm." There's no public evidence this changes how a video is treated.
• "Posting at an exact best time guarantees more reach." TikTok continues recommending videos well after posting, so timing is only one small factor among many.

How the For You feed tests a new video (observation)

Creators commonly describe a pattern where a freshly posted video is shown to a relatively small initial batch of viewers, and how that batch responds — watch-through, shares, comments, re-watches — appears to influence whether the video gets shown more widely. TikTok hasn't published the mechanics of this in detail, so this is a pattern creators report rather than a confirmed process.

If that pattern holds, it would explain why a video can look flat for the first hour and then suddenly pick up pace, or why a video can get a quick early boost and then plateau. Checking a video's performance only in the first few minutes may not tell you much about where it will end up.

Signals that are more and less within your control

How the algorithm likely differs from a simple popularity contest

A system built around individual personalisation doesn't just reward the videos with the most total views — it tries to match each viewer to what they're likely to enjoy. That's a useful way to think about why a niche video with a smaller potential audience can still perform extremely well within that audience, even if its total view count never rivals a broader topic.

Why copying a viral video rarely works the same way twice

Two videos using the same sound, format and even similar captions can perform very differently. Part of the reason is timing — a sound or format can be fresher the first time a particular audience sees it — and part is that the system is responding to each video and each viewer separately rather than applying one fixed rule to a format.

How much a single video can change your trajectory

Because each video appears to get its own round of testing with a fresh batch of viewers, an account that's been quiet for months can still see one video reach far beyond its usual numbers. This cuts both ways: a strong past video doesn't guarantee the next one performs similarly, which is part of why consistency in format and topic tends to matter more than chasing a single past success.

Treat an unusually strong result as useful data about what worked, not as a new baseline you should expect every upload to hit.

Questions worth asking your own analytics

• Which traffic source — For You, search, profile — brings most of the views on your best videos?
• Does average watch time differ meaningfully between your best and worst performing videos?
• Do videos with more comments or shares tend to also have longer watch time, or are these independent in your data?

For practical steps, read How to Get More Views on TikTok and How to Get More TikTok Followers.