TikTok Algorithm Explained: How It Works in 2026 (With Data)

The TikTok algorithm is the recommendation system that decides which videos show up in each person's For You feed. It ranks videos using three kinds of signals: how people interact with them (watch time, rewatches, shares, comments, skips), what the video contains (captions, sounds, hashtags), and, with less weight, language, country and device settings.

Follower count is not one of those signals, which is why a brand-new account can outperform an established one. The bigger surprise is how much popular advice works against you: in Revid's analysis of 176,367 TikToks with 1,000+ likes, videos with 9 or more hashtags had a median of 16K views, against 77K for videos with 4.

#What does "the algorithm" mean on TikTok?

On TikTok, "the algorithm" means the system that builds your For You feed, the personalized feed the app opens on. It is separate from the Following feed, which shows posts from accounts you follow. TikTok's help page on the For You feed explains both feeds and the keyword filters you can apply to them.

Each swipe on For You is a prediction. TikTok guesses which video you are most likely to watch next, shows it, and learns from what you do. Because the feed is built from interests rather than a friend list, a video can reach thousands of strangers who have never heard of its creator.

#How does the TikTok algorithm work?

TikTok scores every video for every viewer with three groups of signals, and it does not weight them equally. Strong signs of interest count more than weak ones: TikTok has explained that finishing a longer video tells it more than, for example, being in the same country as the creator.

The practical reading: settings get a new viewer started, video information tells TikTok who to test your video on, and interactions decide how far it goes. Gelato's guide to the TikTok algorithm walks through the same three groups with more examples if you want a second explanation.

#Which interactions count the most?

TikTok does not publish its weights, so any ranking is an informed estimate. The logic is consistent, though: the more effort an action takes, the more it says about real interest.

1. Watch time and completion. Watching to the end, especially on a longer video, is the clearest sign a video earned attention.
2. Rewatches. A replay means the video was worth seeing again.
3. Shares and saves. Sending a video to someone puts the viewer's own judgment on the line.
4. Comments and follows from the video. Both take more effort than a tap.
5. Likes. Positive, but the cheapest action on the app.

Revid's data shows how scarce the strong actions are. Across the 176,367 TikToks in our study, viewers left 7.1 likes per 100 views but only 0.88 shares and 0.08 comments. A video that earns shares and comments well above those rates stands out from almost everything else in the same feed.

Negative signals matter just as much. Swiping away in the opening seconds, tapping "Not interested" or hiding a creator teaches TikTok to show that kind of video less. If you can only work on one thing, work on watch time: cut every second that does not move the video forward.

#What happens after you post a TikTok?

TikTok has never published fixed traffic allocation rules. Figures like "your video goes to the first 300 to 500 viewers" circulate widely, but they are guesses. What creators can observe, and what matches TikTok's own description of the system, is a loop of testing and expanding.

1. Classification. TikTok reads your caption, hashtags, sound, on-screen text, speech and visuals to work out the topic.
2. First test. The video goes to a small group of viewers likely to care, which can include some of your followers.
3. Evaluation. TikTok compares how that group behaves (watch time, completion, rewatches, shares, comments, skips) with how viewers treat similar videos.
4. Expansion. If signals are strong, the video goes to a larger and broader audience, often including neighboring interests, and the test repeats at that scale.
5. Plateau. When viewers in a new group respond less, distribution slows. A video can pick up again days or weeks later if its topic starts trending or a new audience reacts to it.

Early momentum helps, because a video that collects strong signals quickly moves through more rounds. A slow first hour is not a verdict, though: the test keeps running as long as people keep responding.

#Does the TikTok algorithm care about follower count?

Not directly. TikTok has said that neither follower count nor a history of high-performing videos is a direct factor in recommendations. Larger accounts still get more views on average, simply because more followers see and react to their posts, but a first video from a new account gets tested on the same terms.

That answers a question people often ask: which video app rewards the content itself over existing follower counts? TikTok is the clearest case. ByteDance built the model in China with Douyin, launched in 2016, then acquired Musical.ly in 2017 and folded it into TikTok, keeping the recommendation-first feed. YouTube Shorts and Instagram Reels have since moved toward the same interest-based approach (our breakdown of the YouTube Shorts algorithm covers how Shorts differs).

The same shift explains the rise of micro-influencers: creators with modest followings can build engaged audiences in narrow niches because the feed finds viewers for them.

#What changed in the TikTok algorithm in 2026?

• A retrained US algorithm. TikTok USDS Joint Venture LLC was established on January 23, 2026 to run TikTok in the US, and it will retrain, test and update the recommendation algorithm on US user data. If most of your audience is in the US, retest formats that worked before the change instead of assuming they still do.
• AI labels do not cost reach. TikTok requires a label on realistic AI-generated images, audio or video, and may label uploads automatically when they carry C2PA Content Credentials. TikTok states that the AI-generated label does not affect distribution as long as the video follows the Community Guidelines.
• Payouts follow the For You feed. The Creator Rewards Program pays on qualified views: unique For You feed views, excluding paid, artificial and under-5-second views, and a video starts earning after 1,000 qualified views. Eligibility requires 10,000 followers, 100,000 views in the last 30 days and videos at least one minute long. The algorithm decides your reach and, if you are in the program, your pay.

#How to get more views from the TikTok algorithm

You cannot control the algorithm, but you control every signal it reads. The figures below come from Revid's analysis of 176,367 TikToks with 1,000+ likes posted between October 2025 and October 2026 (the full study of what makes a TikTok go viral). Two limits: the sample only holds videos that already reached 1,000 likes, and the numbers show what strong videos have in common, not proof of cause. For a longer, views-focused playbook, see our guide to the TikTok algorithm for views.

#Open with the payoff, not a preamble

The opening seconds decide whether viewers stay, and staying is the signal that matters most. Openers do not need to be clever: among the 1,588 TikToks with over 100 million views in Revid's index, 51% of those with a transcript open with a plain statement and 16% with a question (see the 100M Club).

Showing often beats telling. In our study, videos with under 40 characters of speech had a median of 123K views and 20.2% passed 1M views, against 68K and 11.9% for videos with talking throughout. Three openers that follow this logic:

• Show the finished result first, then the process: "This used to be the ugliest chair in my house."
• State the claim plainly: "Your plants are dying because of the pot, not the water."
• Start mid-action, with the most interesting frame as the very first frame.

Find more tested openers in our list of TikTok hooks that stop the scroll.

#Use 3 to 4 hashtags, not 10

Hashtags help TikTok classify your video; stacking them adds noise. In Revid's data, 4 hashtags had the best median (77K views, 13.4% past 1M) and 3 hashtags came close at 75K. Videos with no hashtags had a 51K median, and videos with 9 or more dropped to 16K.

Median viewsShare that passed 1M views

20K40K60K80K

77.2K

None123456 to 89+

Posts with 3 or 4 hashtags earned the most views. Captions stuffed with 9 or more hashtags did worst of all, with median views cut by about 80%.

Source: Revid analysis of 176,367 TikToks with 1,000+ likes posted between Oct 1, 2025 and Oct 16, 2026 (61,203 creators)

Show the data

	Median views	Share that passed 1M views
None	51.1K	9.8%
1	64K	11.8%
2	71.3K	12.7%
3	75.4K	13.1%
4	77.2K	13.4%
5	67.3K	12.1%
6 to 8	71.8K	12.7%
9+	15.6K	7.8%

Pick one or two tags that name your niche precisely and one or two broader topic tags, and skip #fyp, which says nothing about the topic. Our page of trending TikTok hashtags by niche shows which tags are rising, and TikTok's Creative Center tracks trending hashtags and songs. A sound also tells TikTok something about your video: use a trending one when it fits the content, not as a bolt-on.

#Write a caption of 41 to 80 characters

Captions are part of how TikTok understands your video. Captions of 41 to 80 characters had the best median in our data (67K views, 12.5% past 1M), while videos with no caption had 36K. That is about one sentence: name the topic in the words people search, then give a reason to comment, such as "Which one would you pick?". A TikTok caption generator gives you options to choose from.

#Make the video as long as the idea, no longer

Length matters less than most advice suggests. In the 12-month study, every length bucket from under 15 seconds to 10 minutes had a median between 58K and 76K views, and 61 to 90 seconds landed at 70K. Videos under 15 seconds had the lowest median (58K) but the most shares: 1.41 per 100 views.

Median viewsShare that passed 1M viewsShares per 100 views

50K100K150K

76.4K

Under 15 s16 to 30 s31 to 60 s61 to 90 s91 s to 3 min3 to 10 min10 min+

Length alone barely moves views: every bucket lands between 58K and 76K median views (the 790 videos over 10 minutes are a small, unusual group). Clips under 15 seconds get shared the most per view.

Source: Revid analysis of 176,367 TikToks with 1,000+ likes posted between Oct 1, 2025 and Oct 16, 2026 (61,203 creators)

Show the data

	Median views	Share that passed 1M views	Shares per 100 views
Under 15 s	57.6K	11.6%	1
16 to 30 s	64.9K	12.3%	1
31 to 60 s	61.8K	11.8%	1
61 to 90 s	70.5K	12.2%	1
91 s to 3 min	63.4K	11.1%	1
3 to 10 min	76.4K	12.4%	1
10 min+	129.6K	13.8%	0

Revid's larger index (about 2.39 million TikToks with 1,000+ likes, all posting dates) found 16 to 30 second videos had the highest median, about 141K, so very short videos once had an edge that has faded for recent posts. Decision rule: if the idea fits in under 15 seconds, keep it there and aim for shares; if you need room to explain, go past one minute, which is also the minimum for Creator Rewards.

#Post when your audience is online, then stop worrying

Timing gives a video a head start, not a result. Among English-language TikToks in our study, posts at 4 a.m. US Eastern had the highest median (120K views) and 7 p.m. posts the lowest (50K), with 3 to 6 a.m. the strongest window. Creators post from everywhere, so those hours also cover daytime in the UK, Europe, India and Asia: treat them as a hint, not a schedule.

#Pre-post checklist

#How to make TikToks from formats the algorithm already rewards

The most reliable way to send strong signals is to start from a structure viewers have already shown they watch. Revid indexes about 2.77 million short-form videos with their engagement data, so the research and the production happen in one place:

1. Search the viral video library by topic, engagement tier and length to find videos that worked in your niche.
2. Run a Deep analysis (free, up to 60 a day): an AI model breaks down the hook, the timed structure and the scroll-stopping moment, with ways to adapt it to your niche.
3. Choose "Make it about your topic" or "Same energy, new script", and Revid writes an original script with new visuals, voice-over, captions and music.
4. Fix any scene in the timeline editor, then publish or schedule straight to TikTok. Revid reports each post's views, likes and comments, so you can remix the ones that did well.

To start from your own script or a link instead, the AI TikTok video generator turns it into a vertical voice-over video.

#TikTok algorithm myths that waste your time

• "Posting at the perfect minute guarantees views." Timing changes the starting audience; the signals decide the rest. A strong video posted at an odd hour still gets tested.
• "#fyp gets you on the For You page." Every video is already a candidate for For You. A generic tag carries no topic information, so it cannot help TikTok find your audience.
• "Deleting a flop hurts your account." TikTok has never said deleting videos triggers a penalty. Remove off-topic videos if they confuse profile visitors, but old videos sometimes pick up views later.
• "Labeling AI content kills reach." TikTok states the AI-generated label does not change distribution for videos within its guidelines. Leaving realistic AI content unlabeled is the real risk.
• "Only the first hour matters." The test-and-expand loop keeps running as long as viewers keep responding.

#How does the TikTok LIVE algorithm decide who sees your stream?

TikTok does not publish how it ranks LIVE streams, but they appear to follow the same idea: show the stream to people likely to care, then widen the audience when viewers stay and interact. Your followers are the first audience, and non-followers can come across your stream in their feed while you are live.

What you can control: a clear topic in the stream title, a quick recap for people who join mid-stream, and regular prompts that turn watchers into commenters. Going LIVE requires being at least 18 and having 1,000 followers, though the follower threshold may vary by region.

#How to reset or retrain your For You feed

Your feed follows your behavior, so you retrain it by changing what you do:

• Long-press videos you don't want and tap "Not interested".
• Add keyword filters for topics you want to see less of; they apply to both the For You and Following feeds.
• Watch, like and follow more of what you want, and swipe past the rest quickly.
• If your app offers an option to refresh your For You feed in your content preferences, use it for a fresh mix; the feed then learns again from what you watch.

#FAQ

#Does reposting on TikTok help the algorithm?

The Repost button shows someone else's video to some of your followers, which can extend that video's reach, but TikTok does not say how much a repost counts as a signal. Re-uploading videos with light edits is different: TikTok does not treat lightly modified reposts as original for Creator Rewards.

#What is a TikTok shadowban?

"Shadowban" is the creator term for a sudden drop in reach with no notice; TikTok does not use the word. The usual causes are guideline violations, sensitive topics or spam-like behavior, which can keep videos out of the For You feed while they stay on your profile, so check your account status in settings for any violations.

#How often should I post on TikTok?

Post as often as you can without lowering quality, on a consistent schedule so you can compare results week to week. More posts mean more tests, but a rushed video that viewers skip sends a weak signal; track your TikTok engagement rate to see whether a higher frequency is helping.

#Why are my TikTok views suddenly low?

Most drops come from the videos, not a penalty: a weaker hook, an off-niche topic or a format your audience has seen many times. Compare watch time and completion for your recent videos against your best ones in TikTok Studio before assuming you are restricted.

#Is the TikTok algorithm the same as YouTube Shorts?

No. Both recommend videos by interest rather than by who you follow, but they are separate systems with different audiences, so test a video that worked on one platform on the other instead of assuming it will perform the same.