TL;DR
- Viewer behaviour drives the feed. TikTok says its For You feed ranks videos using user interactions, video information, and device and account settings, with the last group weighted lowest.
- Finishing a video counts. TikTok names “whether a user finishes watching a longer video from beginning to end” as a strong signal of interest.
- Follower count is not a direct factor. TikTok says neither follower count nor past hits are direct factors, though a bigger audience helps indirectly.
- Leaks and research point the same way. Reporting on an internal “TikTok Algo 101” document described a system optimised for retention and time spent, and independent audits found likes, follows and watching behaviour all shape what people see.
- Most “viral hacks” are folklore. Deleting flops, posting at magic hours or adding a specific kind of face to frame one are anecdotes, not documented ranking rules.
Contents
- What TikTok says its algorithm rewards
- What TikTok says it does not reward
- What leaks and independent research add
- Evidence vs folklore
- What this means for creators and businesses
- How viewers can see and change the system
- FAQ
“How to go viral” advice usually arrives as a list of tricks. The more useful question is what the recommendation system is actually built to reward. TikTok has published its own explanation, an internal document has been reported on, and researchers have tested the feed with automated accounts. Together they give a reasonably clear picture, and it is less mysterious than the hacks suggest: the system tries to predict what a specific viewer will keep watching.
What TikTok says its algorithm rewards
In its 2020 explainer, How TikTok recommends videos #ForYou, TikTok groups the inputs into three families.
| Signal group | What TikTok lists | How much it matters, per TikTok |
|---|---|---|
| User interactions | Likes, shares, accounts followed, comments posted, content created | Main driver |
| Video information | Captions, sounds, hashtags | Helps match videos to interests |
| Device and account settings | Language preference, country setting, device type | “lower weight in the recommendation system relative to other data points we measure” |
TikTok also says signals are not equal. “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.” In plain terms: a video that people watch to the end tells the system more than a video they scroll past after one second, whatever the like count.
Note the direction of these signals. They describe what a viewer does, not what a creator does. The algorithm is not grading your video in the abstract. It is estimating whether each person shown the video is likely to engage with it.
What TikTok says it does not reward
The same post makes two statements that contradict a lot of creator advice.
First, “neither follower count nor whether the account has had previous high-performing videos are direct factors.” TikTok adds that a large following can still raise views indirectly, because more people follow you and engage. So a new account is not locked out by design, but an established one is not starting from zero either.
Second, some content is excluded. TikTok says it does not recommend “duplicated content, content you’ve already seen before, or any content that’s considered spam,” and that videos “that have just been uploaded or are under review” may be ineligible for anyone’s For You feed. That last point explains why a fresh upload can sit at zero views for a while. If that is your main problem, our separate guide on why TikTok videos get 0 views covers the common causes.
TikTok also says the feed deliberately breaks up repetition. It “generally won’t show two videos in a row made with the same sound or by the same creator,” and it mixes in content outside a user’s known interests. A widely repeated claim is that every new video is first tested on a small audience and then pushed wider if it performs. TikTok’s own explainer does not describe such a step, so treat it as a plausible theory rather than a documented rule.
What leaks and independent research add
In December 2021, The New York Times reported on an internal TikTok document titled “TikTok Algo 101”. According to Gizmodo’s summary of that report, a TikTok spokesperson confirmed the document was real and meant for “less tech-savvy employees.” It described the “ultimate goal” as adding daily active users, and named two metrics the system optimises for: “retention” and “time spent.” It showed a simplified scoring formula combining the predicted chance of a like, a comment and a play with expected playtime, and warned that “the equation shown in this doc is highly simplified.”
Independent researchers have tested the feed with automated “sock puppet” accounts:
| Study | Method | Main finding (quoted) |
|---|---|---|
| Boeker & Urman, The Web Conference 2022 | Sock-puppet test accounts varying language, location, follows, likes and watch time | “the follow-feature has the strongest influence, followed by the like-feature and video view rate” |
| Vombatkere et al., 2024 | Analysis of viewing histories donated by real users | The algorithm “exploits users’ interests in between 30% and 50% of all recommended videos”; “liking and following are the primary drivers of personalization” |
| Mosnar et al., 2025 | Attempt to reproduce earlier audits | “these one-shot audit findings often hold only in the short term” |
Two things stand out. Viewer actions (follows, likes, how long they watch) are consistently the strongest levers. And a large share of any feed is exploration: TikTok showing people things it is not yet sure they want. That exploration is where unknown creators get their chance. The third study is a useful warning: the system changes often, so any precise “rule” has a short shelf life.
Evidence vs folklore
The earlier versions of this article, and its companion piece on “rules” for going viral, mixed sensible points with personal anecdotes. Here is how the common tips hold up against the sources above.
| Common tip | Status | Why |
|---|---|---|
| Make videos people watch to the end | Supported | TikTok names full watch-through as a strong signal |
| Captions, sounds and hashtags matter | Supported, with limits | TikTok lists them as video information; they help matching, not ranking on their own |
| Stick to one topic or niche | Plausible | Consistent topics make it easier to match videos to interested viewers; not a stated rule |
| Use popular sounds | Partly supported | Sounds are a listed signal; no source says trending sounds get a boost by themselves |
| You need followers first | Contradicted | TikTok says follower count is not a direct factor |
| Delete videos with 0 views | Folklore | No official or research source supports it |
| Use TikTok’s in-app editor for a boost | Folklore | Not mentioned by TikTok or the studies above |
| A certain face in the first frame boosts reach | Folklore | Unsourced, and not a documented factor |
| Spam hashtags | Counterproductive | Spam and artificially boosted traffic can be excluded from recommendations |
One older point still stands on common sense: trends move fast, and arriving early matters more than arriving with a perfect video. That is an observation about audiences, not a documented ranking factor.
What this means for creators and businesses
If the system optimises for retention and time spent, the practical question for any video is simple: will the people who see it keep watching? A few implications follow.
- The first seconds carry the weight. A video that loses viewers early sends a weak signal, however good the ending.
- Clarity helps matching. Captions, spoken words and on-screen text give the system information about what the video is about, so it can reach people interested in that topic.
- Engagement is earned per viewer. Likes and follows from viewers who actually care are worth more than broad, shallow reach.
- Originality is not optional. Duplicated content is excluded by TikTok’s own description.
For official guidance, TikTok announced an upgraded Creator Academy in March 2024, an education hub with “regularly updated resources, courses, articles, videos, and insights.” Its marketing side, the Creative Center (now listed under the TikTok One Creative Suite), includes a Trends section. Both are better starting points than screenshots of someone’s analytics.
The same logic runs through other recommendation systems, from social feeds to how AI search chooses its sources: systems reward what predicts satisfaction for a specific user, not effort or follower counts.
How viewers can see and change the system
Since December 2022, TikTok has offered a “Why this video” panel, reached through the share menu. The reasons it can show include “user interactions, such as content you watch, like or share, comments you post, or searches,” “accounts you follow or suggested accounts for you,” “content posted recently in your region” and “popular content in your region.”
In the EU, the Digital Services Act requires very large platforms to “provide an option in their recommender systems that is not based on user profiling,” according to the European Commission. In August 2023 TikTok announced it would let European users “turn off personalisation”, in which case the feed shows “popular videos from both the places where they live and around the world.” For creators, this is a reminder that part of the audience may never see a personalised feed at all.
FAQ
Does follower count affect whether a TikTok goes viral?
Not directly, according to TikTok. It says neither follower count nor previous high-performing videos are direct factors, although a larger following can increase views indirectly.
What is the most important signal in the TikTok algorithm?
TikTok does not publish exact weights. It says watching a longer video to the end is a strong signal, and independent audits found follows, likes and view rate among the most influential factors.
Does TikTok test new videos on a small audience first?
This is widely claimed, but TikTok’s own explainer does not describe it. TikTok does say new uploads and videos under review may be ineligible for the For You feed for a time.
Should I delete TikTok videos that get no views?
There is no official or research evidence that deleting low-performing videos improves reach. Low views on a new upload often reflect review or eligibility, not a penalty.
Do hashtags still matter on TikTok?
TikTok lists hashtags as video information that helps recommendations, alongside captions and sounds. They help the system understand a video, but they do not replace viewer engagement, and spammy use can backfire.
Can I turn off the TikTok algorithm?
Not entirely, but in the EU you can turn off personalisation. TikTok offers a non-personalised feed there under the Digital Services Act, showing popular videos from where you live and around the world.
Sources: TikTok Newsroom – How TikTok recommends videos #ForYou · TikTok Newsroom – Learn why a video is recommended For You · TikTok Newsroom – Fulfilling our commitments under the DSA · TikTok Newsroom – Creator Academy update · TikTok Creative Center · European Commission – DSA: very large online platforms · Gizmodo – summary of NYT report on “TikTok Algo 101” · Boeker & Urman (2022), arXiv · Vombatkere et al. (2024), arXiv · Mosnar et al. (2025), arXiv
About the Author
👉Want guest posting?🪙Subscribe to the @egorithms Telegram