Blog
TikTok algorithm: how it picks the videos in your For You feed
The TikTok algorithm estimates, for each person and each candidate video, how likely they are to watch it for a certain amount of time, finish it, like it, share it, comment on it, follow the creator or skip it, and it ranks the For You feed using those predictions. There’s a filter first: only videos that have passed moderation and are eligible for For You are in the running. Then the system swaps out videos that are too similar to each other so the feed doesn’t repeat itself. That’s how TikTok describes it on its recommendation system transparency page.
Neither that page nor the other TikTok pages cited here say how much each prediction weighs or give a formula you can reproduce. If you see percentage weights attributed to the TikTok algorithm, ask where they come from.
What it looks at and what it tries to predict
The transparency page separates three things: information about you, information about the video, and what the system tries to anticipate.
| Type of information | What it includes, according to TikTok |
|---|---|
| About you | Videos you’ve recently interacted with (the ones you watched, finished, liked or skipped), the accounts you follow, your region, your device’s language and its operating system |
| About the video | When and from which region it was posted, the creator’s language setting, its sound, its length and its hashtags |
| What it tries to predict | Whether you’ll like, share or comment on it, or mark it as “Not interested”; whether you’ll follow the creator or interact with their profile; whether you’ll finish it, skip it or add it to your favorites; whether you’ll watch it for a certain amount of time; whether you’ll tap its sound |
It uses those signals to calculate a prediction score for each video and sorts the videos from most to least likely that you’ll enjoy them.
The only official hint about relative weight dates from 2020. In that article, TikTok grouped the signals into your interactions, video information (captions, sounds and hashtags) and device and account settings (language, country and device type), and explained that strong signals carry more weight than weak ones. Its example of a strong signal: someone finishing a longer video from beginning to end. Its example of a weak one: the viewer and the creator being in the same country. Device settings get less weight because they aren’t preferences people actively express. Neither page gives numbers.
Almost everything TikTok tries to anticipate happens while someone is watching your video: whether they stay, reach the end, share it, follow you. Captions and hashtags tell the system what the video is about; what happens while people watch, all the way to the end, feeds the predictions.
Followers and past hits aren’t direct factors
In the same 2020 article, TikTok wrote that neither follower count nor whether the account has had previous high-performing videos are direct factors in recommendations.
The word that matters is “direct.” TikTok also acknowledged that an account with more followers is likely to get more views, because it already has that base. What the system doesn’t do, according to TikTok, is use account size or its last hit as a direct factor. A new account isn’t ruled out for having few followers, and a large account isn’t guaranteed reach on its next video because of what the previous one did.
Why the feed mixes things up
TikTok designs For You not to repeat itself. In 2020 it explained that the feed generally won’t show two videos in a row with the same sound or by the same creator, and that it doesn’t recommend duplicated content, content you’ve already seen or spam. The transparency page adds that if the top-ranked videos for you are too similar to each other, it replaces some of them with others.
The For You feed eligibility standards, in effect since September 24, 2026, add another criterion: the system avoids repeatedly showing content that may lead to a negative experience, such as extreme fitness or dieting, sexually suggestive content, sadness, or overgeneralized mental health advice. During crises or unrest, it may interrupt repetitive recommendations to show a broader mix of content and perspectives.
What makes a video ineligible for For You
The same standards explain that For You is meant for a broad audience, and that content not suited to that audience is left out. The criteria are spread across the Community Guidelines, in the passages that mark what is ineligible for For You. Here are some of them, not the complete list:
| Guidelines section | What is ineligible for For You |
|---|---|
| Integrity and Authenticity | Unoriginal or reused material without anything new. Content that tries to manipulate people into giving gifts or inflating likes or follows. Misinformation that may cause moderate harm. Unverified information about crises or major civic events, or content under review by fact-checkers (as a precaution). Suspected paid political advertising |
| Sensitive and Mature Themes | Significant adult body exposure and sexually suggestive behavior involving adults (also restricted to people 18 and over). Less graphic or public-interest content such as excessive blood, graphic physical altercations, graphic fictional violence or the moments leading up to a serious accident: it may be viewable only by people 18 and over, or be ineligible for For You |
| Mental and Behavioral Health | Activities that may cause moderate physical harm (also restricted to people 18 and over) |
| Safety and Civility | Content that indirectly demeans or furthers the exclusion of protected groups, and lower-severity degrading behavior: the guidelines say it “may” be left out |
Other cases are documented outside that list. The transparency page says content created by accounts under 16 isn’t eligible for For You. The 2020 article notes that a video that has just been uploaded or is under review may not be eligible yet. And if you promote your own business, the Regulated Goods, Services, and Commercial Activities section asks you to disclose it as commercial content: if TikTok finds commercial content that hasn’t been disclosed, it may reduce its visibility. The steps are in how to label commercial content on TikTok.
A video that isn’t eligible for For You still exists: the standards make clear that people can still find it through search or by going to your account. If your strategy also targets those searches, see how to prepare a post for TikTok search.
If your videos aren’t getting views, separate three causes
TikTok’s standards point out that a video with few views didn’t necessarily break a rule. Before you change anything, work out which of these three common causes applies. The order matters, because each step rules out one cause before you move to the next.
- Check whether the video is eligible for For You
Open the video’s analytics in TikTok Studio: according to the standards, that’s where you can see whether a video was made ineligible for recommendation. If there’s a notice, read the reason and, if you disagree, follow how to appeal a video ineligible for the For You feed. If the notice appears below your username and refers to the whole account, the process is different: how to appeal an account recommendation notice. If there’s no notice, go to the next step.
- Make sure you're not measuring too early
A freshly uploaded video may still be under review, and a two-hour-old video can’t be compared with a week-old one. Compare your videos once they’ve all been live for the same amount of time, as explained in how to find your best time to post on TikTok. If you schedule and publish to TikTok with HeyMark, you can also see each post’s metrics there; the eligibility notice, however, is shown by TikTok in its own analytics.
- If it's eligible and has had time, look at where people leave
If it’s eligible, has had time, is your own material and, if it promotes your business, is disclosed as commercial content, the most likely cause is low interest from the people who saw it. TikTok tries to predict whether someone will skip the video, how long they’ll watch and whether they’ll finish it, so the retention curve is the part of your analytics closest to those predictions. Compare it with your similar videos, not your best video of the year, and use the guide to short-form video retention to decide what to edit. If you decide to change it, make a new version instead of uploading the same file again: TikTok says it doesn’t recommend duplicated content.
If the pattern holds after several videos, change one thing at a time and compare across your own posts, as suggested in content testing.
If the change you have in mind is joining a trend, first check how to find TikTok trends and decide whether your brand should join: we explain how to read the curve to tell if you’re still in time, and which music a business can use.
If what you want is to reset your own For You feed
“Resetting the algorithm” can mean two things. If you’re a creator, TikTok doesn’t document an option that resets how your videos are recommended: what changes their path is the videos you post and how people respond to them.
If you’re the viewer and your For You feed no longer looks like what you’re into, TikTok has an option to refresh it. When it announced it in 2023, it explained that it shows the feed as if you had just signed up, that it doesn’t override settings you’ve already turned on or affect the accounts you follow, and that the system then starts recommending based on your new interactions. Neither that announcement nor TikTok’s guide to adjusting your feed says whether it clears your history or which menu it’s in: look for it in the app’s settings.
That guide describes other controls that work without starting over:
- Press and hold a video and tap “Not interested,” so you’re shown less similar content.
- Filter out keywords or hashtags you don’t want to see on For You.
- Choose to see more or less content about specific topics.
- Mute accounts you want to see less of.
- Turn on Restricted Mode, which limits content that may not be comfortable for everyone.
The videos you like or add to favorites, the accounts you follow and what you search for also tell the system what you’re into. Refreshing your feed changes what you see; it doesn’t change how the videos you post are recommended.
Plan your next post in HeyMark.
Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.