What is the TikTok algorithm?
The TikTok algorithm is the recommendation engine behind the For You Page (FYP) — the personalized feed every TikTok user sees on opening the app. Unlike Instagram or LinkedIn, where the home feed leans on people you follow, TikTok's FYP is dominated by content from creators you've never seen before. That's why a brand-new account with the right video can rack up a million views overnight. TikTok officially explained the algorithm in a 2020 blog post and has updated its guidance several times since. The model is closed source, but the core principles are well documented.
What the algorithm looks at
The four pillars of the TikTok ranking system:
- User interactions: Videos you liked, shared, watched to the end, the accounts you follow, the comments you leave, the creators you've hidden
- Video information: Captions, sounds, hashtags, effects, on-screen text — TikTok understands all of these via computer vision and speech recognition
- Device and account settings: Language, country, device type (used for technical optimization, not heavily weighted)
- Negative signals: "Not Interested" taps, fast skips, low watch ratio
The four metrics that matter most
In rank order of impact:
- Watch-through rate: % of the video watched. Over 80% is excellent; over 100% (rewatches) is exceptional
- Completion rate: Did the user watch to the end?
- Replays: How often the video loops
- Shares and saves: Strong proxy for value; comments come next
Likes, perhaps surprisingly, matter much less on TikTok than on Instagram.
How videos are distributed
When you publish a video, TikTok shows it to a small initial test batch (usually a few hundred to a few thousand viewers within your geographic and interest cluster). If those viewers watch through, engage and share, TikTok progressively releases it to larger and larger audiences. Strong videos can stay in distribution for weeks — TikTok is famous for resurfacing old content when patterns suggest it's still relevant.
Working with the algorithm in 2026
- Hook in the first 1–2 seconds: Loud visual change, bold text, intriguing question. If users swipe in those seconds, the video is dead
- Loop the ending: Make the final frame visually identical to the opening so viewers don't realize they're rewatching
- Aim for 21–34 seconds: TikTok's data team has signaled this is the engagement sweet spot
- Use trending sounds: Adds free distribution and signals relevance
- Post 1–3 times per day: More than 4 dilutes
- Engage in the first hour: Reply to early comments to spike engagement velocity
- Niche down: The algorithm rewards creators with a clear topical identity (#booktok, #cleantok, #foodtok)
- Captions on-screen: 85% of TikTok users watch with sound on, but captions still lift completion
Common myths
- "Shadowbans": Most "shadowbans" are simply mediocre videos that didn't break out. True bans only follow community guideline violations
- "Posting at 7pm": Generic peak times exist but vary heavily by audience. Use TikTok Pro analytics
- "Buy a small follower account to boost": TikTok evaluates each video largely independently; account size matters less than on Instagram
TikTok for US small businesses
TikTok now reaches over 150 million US users monthly, with the audience aging up — 38% are now 30+. For brands the practical implications:
- Authenticity wins: Phone-shot, raw videos consistently outperform polished agency content
- Showcase the human behind the brand: Founder-led content is the highest-converting TikTok format
- TikTok Shop is now embedded for US merchants: link products directly to a video
- Cost-effective ads: Spark Ads (boosting organic posts) deliver some of the lowest CPMs in social right now ($5–$10 in 2025)
Common mistakes
- Cross-posted Reels with watermarks: TikTok dramatically suppresses these
- Polished commercials: Feel out of place on the platform
- Aggressive hard-sell: Doesn't match platform behavior; lowers watch time
- Inconsistent posting: Algorithm rewards steady cadence
- Ignoring trending audio: Free distribution left on the table
- No on-screen text: 85% of users keep sound on, but on-screen text still improves completion
How TikTok evaluates account quality
TikTok keeps a hidden "creator quality score" that affects baseline distribution. Signals that lift it:
- High average watch-through across recent videos
- Consistent posting cadence
- High follower-to-view ratio
- Original content (not reposted)
- Low rate of community guideline violations
Signals that lower it:
- Frequent low-watch-time videos
- Long inactive gaps
- Repeated minor violations
- Watermarked content from other platforms
This is why one "viral" video on a quality account triggers many viral videos in a row, while the same video on a low-quality account often plateaus.
publy.ch helps US small businesses produce TikTok and Reels content at scale — hook-first visuals, captions, on-screen text — designed for the algorithm signals that currently matter most.