The Hook Mine
There is a method going round: find ten competitors with fifty thousand followers, open their reels, filter by Most Viewed, save the top twelve from each, then steal the hook word for word and change the rest.
The middle step is the good idea. The first and last steps are where it quietly goes wrong — one of them ranks by follower count instead of by hook, and the other hands you a sentence you do not own, on the same day it hands the identical sentence to everyone else who watched that video. This page runs the fixed version live, in your browser, on hooks you paste in yourself.
Most Viewed is a follower sort, not a hook test.
On an account with 4.9 million followers, the twelve most-viewed reels are mostly the ones the algorithm decided to push. Copy those hooks and you have copied the distribution, not the writing. The only honest question is: did this post beat what this account normally does? A 96k-view clip on a page whose median is 11k is a hook that worked. A 2.4 million-view clip on a page whose median is 2.3 million is a Tuesday.
Their sort — most viewed
Ours — outlier vs its own author
Keep the skeleton. Never keep the sentence.
A hook that works has a shape: a condition, a timeframe, a promise of a plan. Shapes belong to nobody — they are older than the platform. The words inside the shape belong to the person who wrote them, and the moment forty accounts in one category publish the same twelve sentences, the sentences stop working for all forty. So the mine keeps the structure, recasts it in your own claims, and then measures what survived: the longest run of their meaning-carrying words, and how much of their vocabulary you reused in any order. Over the line, it is blocked and you are shown the offending run.
Run on the two reels that started it.
Both of these landed in our own feed. Both were watched frame by frame and transcribed before a single line of this was written. Here is what the mine returned on them — including the one place it blocked our own pipeline, because a skeleton it could not classify had nothing new to say and tried to hand back the source sentence.
| Source hook | Outlier | Archetype | Ours | Gate |
|---|---|---|---|---|
| I helped a local business replace its entire marketing team with an agentic AI a builder's 2-minute reel | 8.73× NUCLEAR | Result receipt | I helped a two-room clinic to answer every after-hours form in under a minute | CLEAR 1 content word in a row, 6% vocabulary |
| If I wanted to create four months of content in the next thirty minutes, this a coach's 26-second reel | 3.42× PROVEN | Conditional method | If I wanted to fill a med spa's injector calendar in one afternoon, this is exactly what I would do | CLEAR 1 content word in a row, 9% vocabulary |
| Five mistakes to avoid when you scale a 9-figure account in the same category | 1.04× ORDINARY | Numbered method | Four mistakes to quietly empty a clinic's Tuesday | CLEAR 1 content word in a row, 11% vocabulary |
| Stop posting transformation photos a coach's 26-second reel | 1.0× ORDINARY | Mistake / stop-doing | Stop answering leads the next morning — it is costing you the Tuesday chair | CLEAR 1 content word in a row, 10% vocabulary |
| Why your automation breaks every time a tool updates? a builder's 2-minute reel | 0.73× ORDINARY | Question gap | Why does a paid lead go cold in nine minutes? | CLEAR 0 content words in a row, 9% vocabulary |
| Watch me wire a speed to lead responder in one sitting a builder's 2-minute reel | 1.0× ORDINARY | Proof demo | Watch a Sunday-night form fill turn into a Monday appointment | CLEAR 1 content word in a row, 8% vocabulary |
| In ninety days I doubled my retainer a 9-figure account in the same category | 0.96× ORDINARY | Timeline | In one quarter I rebuilt a clinic's front desk around a text thread | CLEAR 0 content words in a row, 0% vocabulary |
| Everyone tells you to niche down. It is the slowest way to start. a 9-figure account in the same category | 1.0× ORDINARY | Contrarian | Everyone tells you to post more. Nobody books from volume | CLEAR 2 content words in a row, 20% vocabulary |
| Three client-getting scripts to fill your roster a coach's 26-second reel | 0.75× ORDINARY | Unclassified | — nothing new to say, so nothing shipped | BLOCKED 6 of their content words in a row: "three client getting scripts to fill your roster"; 100% of the meaning-carrying words reused (ceiling 45%) |
Outlier is views divided by that account's own median across the clips mined, not by its follower count. The gate figures are the longest run of meaning-carrying words shared with the source and the share of its content vocabulary reused in any order.
A calendar, not a folder of screenshots.
A hundred and twenty saved videos is a folder. What a business needs is a sequence: the same archetype never twice in a week, the same source account never twice in five days, and a funnel mix that is mostly attention but never only attention. The mine emits the sequence and a brief per day, and it reports its own recycle factor — if you fed it three competitors instead of ten, the calendar says so rather than pretending to a hundred and twenty original days.
The awkward ones.
Is it wrong to study a competitor's hooks at all?
No, and anyone who says otherwise has never written for an audience. Structure is how communication works — the numbered list, the conditional promise, the callout are older than every platform they run on, and no one owns them. What you cannot take is the sentence somebody wrote, because that is theirs and because the moment forty accounts in one category publish the same twelve sentences, the sentences stop working for all forty.
Why is sorting by Most Viewed actually wrong?
Because view count is dominated by the size of the account and what the algorithm decided to push, not by the quality of the first line. On a five-million-follower page the top twelve reels are the ones distribution favoured. The measure that tells you the hook did something is the multiple against that same account's own median — and on that measure a 96,000-view clip on a small page routinely outranks a 2.4-million-view clip on a huge one.
What does the gate actually measure?
Two things, because either alone is easy to fool. First, the longest run of meaning-carrying words you share with the source — scaffolding like 'this is exactly what I would do' is not counted, because that is the shape, not the claim. Second, how much of their content vocabulary you reused in any order, which catches the shuffled paraphrase that never repeats four words in a row but is still their line. Over either ceiling, it is blocked and the offending run is printed.
Does a plagiarism check make the hook worse?
It makes it different, which is the point. In our own run the source hook and our recast share the same archetype and one meaning-carrying word, and the recast is about a clinic's Tuesday instead of four months of content. The structure that earned the attention survives; the sentence that would have flagged you does not.
What happens when nothing in the category is worth mining?
It says so. Load the preset where every account is publishing at its own baseline and the mine returns no proven hooks at all, because none of them beat their author. That is a real and useful answer — it means the category is flat and the opening is a format nobody there is running, not a better version of what they already do.
Do the hooks get checked against what my business may legally claim?
Yes. A second gate reads every generated line against the brand's claims registry and kills guarantees, medical verbs, superiority claims and earnings promises that the business cannot substantiate. A hook that would get a clinic in trouble is not a good hook however well it performs on someone else's page.
Can you run this on my category?
Yes, free. Send the ten accounts you keep losing to. You get the ranked hooks with their outlier multiples, the recast versions in your own claims, the sequenced calendar — and the blocked list, published next to it, so you can see exactly what we refused to hand you.
Want your category mined?
Send the ten accounts you keep losing to. We come back with the hooks that actually beat their own authors, recast in your claims, every one of them through the gate — and the blocked list published next to the calendar.
Text Rudy →