How TikTok Algorithm Works

Have you ever scrolled through TikTok and caught yourself thinking, "I was looking for how to draw eyeliner, and now they're showing me a perfect biscuit recipe"? Or you liked a cat video, and an hour later your whole feed is about Stoic philosophy?

Welcome to the reality of 2026.

The question of how the TikTok algorithm works is the most popular among bloggers, marketers, and ordinary users. The platform's official stance sounds smooth: "We show you what you're interested in." But the truth is more complex. TikTok's algorithms are a neuro-psychotherapist that knows you better than your mother.

Why does watching makeup videos lead to cooking hacks? Why is a gamer shown crocheting, and an accountant — house construction?

Today, we'll open the "black box" of the recommendation system. We'll dissect 7 layers of ranking and prove: the TikTok algorithm doesn't just select content — it reassembles your personality from scratch.

Part 1. Anatomy of the algorithm: from "For You" to "This is You"

To understand how the TikTok algorithm works, forget everything you know about Instagram and YouTube. They work with social graphs (who follows whom). TikTok has cognitive graphs. It doesn't care about subscriptions. It only cares about one thing: what your brain does with videos.

Three basic layers of TikTok recommendations:

User signals — what you actually did: watched until the end, looped, forwarded to a friend, clicked "not interested."
Video information — hidden captions, audio track, hashtags, objects in the frame (AI recognizes them automatically).
Device context — interface language, geolocation, phone model, time of day.

But the main difference between TikTok and YouTube is the time factor. YouTube remembers what you loved 5 years ago. TikTok remembers what you loved 5 minutes ago and instantly reconfigures the feed.

It is precisely because of this constant re-learning that the phenomenon arises: you clicked on a makeup video, didn't watch it all the way (boring), then liked a food video (tasty). The algorithm concludes: "This user finds makeup boring, but food interesting. Let's remove cosmetics and add pizza."

But this is a simple case. Much more interesting is the deep connection.

Part 2. The "neighboring neurons" effect: why makeup and cooking are relatives

Why does watching makeup videos lead to culinary hacks from the perspective of mathematics and neural networks?

The answer lies in cross-domain clustering. TikTok's neural network breaks down each video into micro-patterns. It doesn't distinguish mascara from flour. It distinguishes a behavioral pattern.

What the algorithm sees in a makeup video:

Visual cues: close-up of a face, hands, brushes, mirror, directed light
Auditory cues: whispers, ASMR sounds of application, instrumental music without words
Behavioral cues: instruction ("first base, then foundation"), "before/after" transformation, focus on details

What the algorithm sees in a culinary hack video:

Visual cues: close-up of hands, bowl, spatula, loose ingredients, mixing process
Auditory cues: sizzling oil, knife tapping on a board, calm voice-over
Behavioral cues: instruction ("first flour, then eggs"), "dough/pie" transformation, focus on tactile sensations

Do you see the coincidences?

The algorithm concludes: "The user likes to watch the process of creating a result with their own hands." And it gives you the next similar process — cooking.

Other examples of non-obvious connections:

You watched a blogger's wedding → the algorithm decided: you like social rituals and following traditions → you will be shown: morning coffee ritual, religious ceremonies, scheduled exercises.
You got hooked on apartment renovation → the algorithm decided: you enjoy the visual transformation of space → you will be shown: cleaning, wardrobe rearrangement, before/after desk organization.
You watch a dance challenge → the algorithm decided: you like rhythm and repetitive cyclical actions → you will be shown: sports, boxing, Rubik's cube assembly, video editing to a beat.
You are learning to draw eyeliner (makeup) → the algorithm decided: you love morning self-setup rituals → you will be shown: breakfast preparation, a glass of lemon water, morning affirmations.

That is, the TikTok algorithm connects videos not by topics, but by emotional and behavioral scenarios.

Part 3. Deep learning: how the algorithm predicts your desires before you do

You think you control your phone. But how does the TikTok algorithm really work? It uses Reinforcement Learning.

Every action you take is a reward or punishment for the neural network.

Action weighting system (based on analysis of TikTok patents 2024–2025 and leaks from former engineers):

Watching a video to the end → +10 points. The strongest signal. You're hooked.
Rewatching / looping (2+ times) → +50 points. The algorithm is ecstatic. You're addicted.
Like → +5 points. Good, but weak. Likes are often given out of inertia.
Comment → +15 points. High-order engagement.
Repost / sharing → +20 points. Social validation. The algorithm loves it.
Scrolling past (without stopping) → −5 points. Bad.
Clicking "Not interested" → −100 points. A nightmare. The algorithm will remember this topic for weeks.

Now back to our makeup example.

You watch a makeup video. The scenario is boring. You scroll halfway and leave. The algorithm receives a signal: "Makeup in this particular pattern did not retain the user."

After 3 videos, you see a video of a guy mixing pizza dough to ASMR. You watch it to the end. Then again. Like.

The algorithm makes a note: "The user likes close-ups of hands + instructions + tactile ASMR."

Deposit funds, one-click order, discounts and bonuses are available only for registered users. Register.

If you didn't find the right service or found it cheaper, write to I will support you in tg or chat, and we will resolve any issue.