AI TikTok Analyzer Pro

2026-08-16 · AI TikTok Analyzer Pro Team

9 TikTok Hook Formulas (and How to Find Which One a Video Used)

EnglishEspañolBahasa Indonesia日本語한국어PortuguêsTiếng Việt

A hook is a promise made in the first two or three seconds. There are not infinite ways to make one — there's a small set of structures that keep reappearing because they each solve the same problem differently: giving a scrolling stranger a reason to stay.

Below are nine that recur across categories, with the template and the way each one fails. Then, more usefully, a method for taking any video that worked and identifying which structure it used — because copying a specific line rarely transfers, and copying the structure usually does.

The nine at a glance

#FormulaCore move
1The Specific NumberPromises a countable payoff
2The NegationContradicts an assumed belief
3The QualifierNames one audience out loud
4The Open LoopWithholds the payoff
5The CredentialBorrows authority up front
6The ConfessionTrades a mistake for trust
7The Cold OpenLets the frame do the work
8The Standing QuestionSays the thing viewers already wonder
9The ComparisonPuts two states next to each other

1. The Specific Number

Template: "[N] [things] that [outcome]" — "Four settings I change on every new phone."

Numbers work because they bound the commitment. The viewer knows how long this will take and what they're getting.

Failure mode: round, large numbers. "50 tips" promises a list nobody finishes. Small and odd reads as researched rather than padded.

2. The Negation

Template: "Stop [common practice]" / "[Common belief] is wrong."

Contradiction creates a gap the viewer wants closed, especially if they do the thing you just called out.

Failure mode: negating something nobody believed. If the "myth" is a straw man, the video reads as manufactured and the comments will say so.

3. The Qualifier

Template: "If you [specific situation], [promise]" — "If you rent and can't drill holes, this one's for you."

You deliberately shrink the audience. The people who match feel addressed rather than advertised at, and they're the people who'd convert anyway.

Failure mode: qualifying too broadly. "If you use your phone" isn't a qualifier; it's noise.

4. The Open Loop

Template: "[Setup] … and then [withheld]."

An unresolved question is uncomfortable, and the discomfort buys you seconds.

Failure mode: the loop never closes, or closes on nothing. "Wait for the end" followed by an anticlimax trains the audience to distrust you, and that lesson persists across videos.

5. The Credential

Template: "I've [done X] for [N years] — here's [claim]."

Authority compresses the trust-building that would otherwise take the whole video.

Failure mode: credential without specificity. "As an expert" is weaker than "I've fitted about 200 of these." Also: credentials imply claims, and claims in regulated categories — health, finance, safety — carry obligations. Don't borrow authority you can't back.

6. The Confession

Template: "I [made this mistake] so you don't have to."

Admitted error is disarming, and mistakes carry information that advice doesn't.

Failure mode: the fake confession. A "mistake" that turns out to be a humblebrag is transparent, and audiences call it out in the comments faster than anywhere else on the internet.

7. The Cold Open

Template: no opening line at all — an arresting frame, a physical action mid-motion, an unexplained object.

This is the one people forget is a hook. The transcript is empty; the work happens visually.

Failure mode: a striking image with no connection to what follows. Bait earns the stay and loses the follow.

8. The Standing Question

Template: "Why does [annoying universal thing] always [happen]?"

You articulate a thought the viewer has had and never voiced. Recognition is a strong reason to stay.

Failure mode: asking a question the audience isn't actually asking. This is the formula most improved by comment research — the questions people genuinely ask are already written down under your competitors' videos.

9. The Comparison

Template: "This vs. this" / a before-and-after held in one frame.

Contrast is legible in under a second, with or without sound.

Failure mode: an unfair comparison. Rigged before-and-afters get read as ads instantly, and in some categories they're an advertising-standards problem, not just a credibility one.


How to find which formula a video used

Watching a video and guessing gives you a feeling. Reading its first sentence as text gives you an answer. Here's the pass.

Step 1 — Build the candidate set

Don't analyze random videos. Analyze videos that outperformed their own account's baseline, which is a different set from "videos with big view counts."

TikTok's profile grid offers reverse-chronological posts and a coarse Popular tab, with no sort by exact plays, likes, comments, or date. Sort the profile properly with our free TikTok video sorter, eyeball the median of the last twenty posts, and flag anything clearly above it. Ten to fifteen videos is plenty.

Step 2 — Get the words out of the video

Pull each one as text with our subtitle downloader. Choose SRT when you want timestamps — you specifically want to see what was said before the 3-second mark — and TXT when you just want the copy. AI transcription also covers videos with no uploaded captions, which is most of them.

Then paste the first sentence only of each video into one document, one per line. Ten hooks stacked in a list reveal patterns that ten viewings never will.

Step 3 — Classify with four questions

Work down your list and ask, in order:

QuestionIf yes
Is there a number in the first line?Formula 1
Does it contradict a belief, or admit an error?Formula 2 or 6
Does it name a specific situation or audience?Formula 3
Does it promise something it doesn't yet deliver?Formula 4

If none apply, the hook is probably 5, 7, 8, or 9 — check for a credential claim, an unanswered question, an implied comparison, or no speech at all.

Step 4 — Check the visual layer separately

This is where transcript-only analysis misleads. Re-watch the first three seconds with sound off. If the video still holds without words, the real hook is visual — a Cold Open or a Comparison — and the spoken line is doing less work than the transcript suggests. Note both layers; strong videos usually run one of each.

Step 5 — Get a second opinion

Run the same videos through the hook analyzer, which classifies the opening, and the script analysis tool, which maps the whole structure — where setup ends, where the payoff lands, how the close is handled.

The point isn't to outsource the judgment. It's the disagreements: where the tool reads a hook differently than you did, one of you missed a layer, and figuring out which is the most instructive minute of the exercise.

Step 6 — Keep a swipe file

One row per video, and revisit it before every script:

VideoFormulaFirst line, verbatimVisual layerWhy it worked here

After thirty rows you'll notice your niche leans on two or three formulas and ignores the rest — which is either a convention worth respecting or a gap worth taking.

Common mistakes

Copying the line instead of the structure. A recycled hook line reads as recycled. The transferable part is the shape: a specific number, a named situation, an admitted error.

Analyzing only monster hits. The biggest video on an account is often carried by a sound or a duet rather than its hook. The third through eighth outliers are more repeatable.

Ignoring the on-screen text. Many hooks are typed, not spoken. Transcripts won't always capture burned-in text, so read the frame as well as the SRT.

Assuming a formula "performs best." Which structure works depends on category, audience, and what that feed has been served all week. Anyone quoting a universal retention figure per hook type is guessing. Test within your own account and compare against your own baseline.

Stacking all nine. A number, plus a negation, plus a qualifier, plus an open loop, in three seconds, is not a hook — it's an auction.

Studying hooks and skipping the payoff. A great opening on a video that doesn't deliver damages the account. Structure the whole thing, not just the front.

FAQ

How long is a hook, exactly? Treat the first two to three seconds as the decision window, and the first sentence as the unit of analysis. Don't over-engineer the boundary.

Can I identify a hook without a transcript? You can guess. Reading ten first lines side by side as text is what makes the pattern obvious — and it takes less time than watching them.

Do these formulas apply to Reels and Shorts? The structures are general to short-form video. The tooling described here works on public TikTok pages.

Does AI transcription handle accents and slang? Reasonably well for structural analysis. For anything you'll quote publicly, check the transcript against the audio — names, slang, and heavy background music are where errors cluster.

**Where do I get ideas for what to hook about?** From comments. The questions and objections under videos in your category are pre-written Formula 8 hooks; our comment exporter pulls a thread into a spreadsheet so you can sort them and pick.


Disclosure: This guide is published by the team behind AI TikTok Analyzer Pro — a browser extension for Chrome and Edge plus a web workbench at tiktok.poviai.com, with a free quota and paid plans at $19.9 and $49 per month. It works on public TikTok pages. It is an independent product, not affiliated with, endorsed by, or sponsored by TikTok or ByteDance. TikTok is a trademark of ByteDance, referenced for identification only. Download and reuse only content you are authorized to use.

Research public TikTok pages with sorting, downloads, transcripts, translation and AI analysis — free to start. Add to Chrome Try the free web tools