TikTok algorithm how it works
How does the TikTok algorithm work and how does it decide what I see?
Projekt-Plan
{{whyLabel}}: TikTok's primary innovation is prioritizing content relevance over who you follow, which is why you see strangers on your feed.
{{howLabel}}:
- Recognize that TikTok builds a 'persona' for you based on micro-interests (e.g., #MechanicalKeyboards instead of just #Tech).
- Understand that your follower count is not a primary ranking factor for distribution.
- Note that the algorithm treats every video as a standalone unit of content.
{{doneWhenLabel}}: You can explain why a user with zero followers can still get 1 million views.
{{whyLabel}}: Knowing which data points the algorithm values most helps you understand why certain videos dominate your feed.
{{howLabel}}:
- User Interactions (70%): Focus on watch time, shares, and saves.
- Video Information (20%): Focus on captions, sounds, and on-screen text keywords.
- Device/Account Settings (10%): Note that language and location are only baseline filters, not growth drivers.
{{doneWhenLabel}}: You have memorized the three main signal categories and their relative importance.
{{whyLabel}}: Retention is the single most important metric for the TikTok algorithm in 2025.
{{howLabel}}:
- Completion Rate: The percentage of people who watch to the very end.
- Re-watch Rate: The strongest signal; if users watch twice, the algorithm considers it 'highly engaging'.
- Average Watch Time: Aim for at least 50% of the total video length to stay in the recommendation loop.
{{doneWhenLabel}}: You can identify the 'hook' and 'payoff' structure in 5 consecutive FYP videos.
{{whyLabel}}: In 2025/2026, the algorithm weights 'Saves' and 'Shares' significantly higher than 'Likes'.
{{howLabel}}:
- Saves: Signal that the content is useful or worth revisiting (e.g., tutorials, recipes).
- Shares: Signal that the content is socially relevant or relatable.
- Comments: Signal 'session depth'; the algorithm rewards videos that start conversations.
{{doneWhenLabel}}: You understand why a video with 1,000 saves often outperforms a video with 10,000 likes.
{{whyLabel}}: TikTok has evolved into a search engine; the algorithm uses keywords to categorize and serve your video to the right 'search intent'.
{{howLabel}}:
- Place keywords in the first 3 seconds of on-screen text.
- Use descriptive keywords in the caption (avoiding generic hashtags like #FYP).
- Speak your keywords clearly; the algorithm transcribes audio to understand the topic.
{{doneWhenLabel}}: You have identified 3 keywords for a specific niche using the TikTok search bar's 'suggested' terms.
{{whyLabel}}: TikTok does not show your video to everyone at once; it uses a tiered testing system.
{{howLabel}}:
- Tier 1 (The 300-User Test): Your video is shown to a small, diverse cohort to gather initial data.
- The Point System: Aim for a score of 50+ points (Likes=1, Comments=2, Shares=3, Completion=4, Re-watch=5).
- Tier 2 & 3: If the score is high, the video is pushed to 10k, then 100k+ users.
{{doneWhenLabel}}: You can explain the 'batch testing' logic and why views often 'stall' at 200-300.
{{whyLabel}}: The algorithm automatically suppresses content that violates safety guidelines or is 'low quality'.
{{howLabel}}:
- Avoid 'borderline' content (e.g., dangerous stunts, excessive skin) which triggers a 'shadow-throttle'.
- Ensure high-quality lighting and audio; the AI penalizes blurry or silent videos.
- Check for 'unoriginal content' flags; reposting videos from other platforms with watermarks will kill reach.
{{doneWhenLabel}}: You have reviewed the 'Account Status' section in TikTok settings to check for violations.
{{whyLabel}}: To understand the algorithm, you must see how it reacts to your own behavior.
{{howLabel}}:
- Scroll for 10 minutes and note why each video was shown (e.g., 'Because you liked a video from [Creator]').
- Identify which videos you 'bounce' from (scroll past in <2 seconds) and see if the algorithm stops showing that niche.
- Check your 'Watch History' in settings to see the data TikTok has collected.
{{doneWhenLabel}}: You have listed 5 distinct 'interest clusters' the algorithm has assigned to you.
{{whyLabel}}: You can manually reset or fine-tune the algorithm if your feed becomes repetitive or irrelevant.
{{howLabel}}:
- Long-press on a video and select 'Not Interested' to remove a specific niche.
- Use 'Refresh your For You feed' in Content Preferences to start from a clean slate (like a new account).
- Set up Keyword Filters in settings to block specific topics from appearing entirely.
{{doneWhenLabel}}: You have successfully filtered out one unwanted topic from your feed.