Nobody outside ByteDance knows TikTok’s ranking function. But TikTok documents more of it than any other major platform, which makes this one of the few algorithm pages where most of the content can be sourced first-party.
The headline, in TikTok’s own words: recommendations are based on user interactions, video information such as captions, sounds and hashtags, and device and account settings, all “weighted based on their value to a user.”
What TikTok says it optimises for
TikTok states that a strong indicator of interest, “such as whether a user finishes watching a longer video from beginning to end, would receive greater weight than a weak indicator, such as whether the video’s viewer and creator are both in the same country.”
Two exclusions are stated outright. TikTok says that neither follower count nor whether an account has had previous high-performing videos is a direct factor in the recommendation system. That is a stronger claim than any other platform makes, and it is the reason a first video can outperform a hundredth.
TikTok also describes deliberate diversification: your For You feed “generally won’t show two videos in a row made with the same sound or by the same creator.”
Note: Figures here were verified against TikTok’s Newsroom explainer “How TikTok recommends videos #ForYou” and TikTok’s Transparency Center recommendation system page as of September 2026. Platforms change these without notice.
What signals does TikTok rank on?
Ordered as TikTok presents them.
| Signal group | Examples TikTok names | Stated weight |
|---|---|---|
| User interactions | Videos you like or share, accounts you follow, comments you post, content you create | Highest, and completion of a longer video is called out as a strong indicator |
| Video information | Captions, sounds, hashtags | Used for matching |
| Device and account settings | Language preference, country setting, device type | Explicitly described as lower weight, included for performance not preference |
| Follower count | Not a direct factor | Stated as excluded |
| Past video performance | Not a direct factor | Stated as excluded |
What changed most recently
Stated by TikTok: the recommendation explainer above is the live published position, and TikTok’s in-app “Why this video” feature surfaces a per-video reason drawn from the same signals.
Widely reported, not confirmed: creators consistently describe a staged rollout, where a video is shown to a small batch and promoted further if retention holds. TikTok has never published batch sizes or thresholds. The general shape is plausible and matches how recommender systems work. The specific numbers circulating (500 views, then 2,000, then 10,000) have no source.
Practitioner heuristic: watch time as a percentage of video length is treated as the primary lever, which is why very short videos that loop tend to over-index. TikTok has confirmed completion matters. It has not confirmed a length strategy.
Myths this kills
Shadowbans. TikTok has no confirmed mechanism by that name. It does have documented ineligibility for the For You feed under its Community Guidelines, and content that is ineligible is genuinely not recommended. The difference matters: one is a rumour, the other is a policy you can read and check. See our note on what a shadowban actually is.
Engagement pods. Pointless here, and more clearly pointless than anywhere else. TikTok states follower count and prior performance are not direct factors, and the feed is built on per-viewer prediction. A coordinated like ring gives the model nothing to generalise from.
“The algorithm hates links.” Nothing published supports this on TikTok. Off-platform destinations are governed by policy, not by a ranking penalty on the presence of a URL.
Posting time is decisive. Less true on TikTok than anywhere else, because content is not primarily distributed to followers in reverse-chronological order and videos routinely pick up views days later. Our best time to post on TikTok page treats the windows as a starting point, and the TikTok posting time tool converts them to your timezone. Volume matters more, which is the subject of how often to post on TikTok.
What actually moves distribution, in priority order
- Retention through the first three seconds. Nothing else gets a chance if this fails.
- Completion rate, and re-watches. The one thing TikTok explicitly calls a strong indicator.
- Shares. A share exports the video to a new audience graph entirely.
- Comments, especially replies to comments. Cheap for the viewer, and they extend the life of the video.
- Sound and topic clarity. TikTok names captions, sounds and hashtags as the matching layer. This is how the system knows who to try it on.
- Volume. Because follower count is not a factor, each video is close to an independent draw. More draws, more chances.
What to do
Front-load the payoff. Keep the video only as long as the idea holds. Give people a reason to send it to someone. Then post again, because on TikTok the previous video’s performance is stated not to help or hurt the next one.
Scheduling TikTok alongside everything else
Volume is the lever TikTok’s own documentation implies, and volume is a workflow problem. BulkPublish queues TikTok videos alongside Reels and Shorts from one calendar, so the same cut goes out in three places without three separate uploads. The bulk publishing feature page covers uploading a batch at once.
The parts you control
You control the format, the hook, and whether you keep publishing. TikTok’s own explainer removes the two excuses people lean on most: your follower count is not holding you back, and your last video did not poison this one. What is left is whether the first three seconds work.