I scraped 193,355 hooks for this.

The hook everybody makes is losing.

I’ve posted it dozens and dozens and dozens of times. “Here’s how to do X.” Teach something useful, post it, move on.

It feels like the safe call. If you run a brand, it’s probably the one your team greenlights most.

I’m guilty too. I wrote an X article called “How To Run Your Meta Ads With Openclaw” (Openclaw is the AI agent that runs my ads). It did 32,819 views.

Same topic, different hook: “I run my meta ads with @openclaw for $0/month.” That one did 632,000.

Different formats, so not a clean test. But that gap stuck with me.

So a couple weeks ago I had an AI read the hooks on 193,355 short videos from TikTok and Instagram. Every one. Then I checked which videos actually took off.

The #1 most-repeated hook formula in the whole pile was that one. The lesson. “Here’s how to do X.” 3,175 videos, more than any other.

It does worse than average. Make the lesson about saving time (“do X 10x faster”) and it takes off about a quarter less often than the average video.

Every one of those videos starts with worse odds before anyone watches a second. If you pay creators or editors per video, that’s budget walking straight into the weakest lane.

I call it the how to tax.

I’m calling the whole project the Hook Atlas. I turned the first round into a free 12 slide report, plus a 3-page field guide. Link’s at the bottom.

The AI I used is called Jev. It’s a new model from TypeSafe that reads text and answers questions about it, fast and for pennies.

Last month I had it break down 723 live ads for 9 cents. That demo did over a million views on X.

So I went bigger.

I pulled 193,355 videos over a weekend. Jev read the words on screen and the caption of each one and answered 22 questions about it. What’s it promising? Who’s it for? What feeling is it going for?

Three of those questions do most of the work. Here’s one real hook, taken apart:

→ what’s in it for me? a discovery: a fact I didn’t know

→ why do I care? my health

→ what do I see first? a shocked “wait…?!”

Every hook makes those three choices. I call the three together a combo.

(Jev also tagged the one feeling each video goes for, like fear, hope or humor. That shows up later.)

Here’s the part I like most. Jev never saw the views. Not the likes, not the follower count. So a viral video couldn’t trick it into thinking a hook was smart.

Then I compared each video to the same account’s typical video (the middle of the pack for that account). If it got 3x the views of that typical video, it broke out.

Why compare an account to itself? A celebrity’s worst video beats most people’s best. This way a 5,000-follower account and a 5-million-follower account get judged the same.

About 1 in 5 videos broke out. 18.7% (of the hooks I scraped, expect the average breakout percentage to be far lower). That’s the bar for everything below.

For scale: spend 30 seconds on each hook and you’re at 1,611 hours. About 40 work weeks of someone doing nothing else.

This took a weekend.

Here’s what came back.

→ the lesson formula (“here’s how to do X,” aimed at getting better at something): 16.7% broke out across 3,175 videos. Below the 18.7% bar.

→ the same lesson aimed at saving time (“do X 10x faster”): 13.6% across 2,113 videos. About a quarter less often than average.

→ any video whose big reason to care is saving time: 15.4%, the weakest reason in the whole pile

And my first reaction… yeah that feels right.

Because I’ve been betting on this for a while. I call it puppet strings.

Every piece of content should pull one string. One deep human want: wealth, health, status, romance, escape. The hook, the examples, the CTA, all pulling the same string.

We’ve run it for brands and creators. It’s a big part of the system behind one client’s 11M views in under 28 days. (Virlo, the video research tool we use, wrote up how we built it here.)

But it was built on my gut and a few client wins. I’d never tested it against hundreds of thousands of videos.

So I checked my strings against 193,355 videos:

→ wealth (money): 24.4% broke out

→ health: 21.1%

→ romance and belonging: 18.8%

→ status: 18.2%

→ escape: 17.0%

→ saving time: 15.4%

Saving time isn’t even one of my strings. It came in dead last.

You want to share knowledge. Teaching people makes sense. But you can’t just teach so they know. You have to connect it to a higher emotion.

Pull a string. Money you could make or lose. Your health. Being wanted. The stuff people actually lie awake about. Nobody lies awake about saving 10 minutes.

And because so many people run the how-to hook, it’s harder to pop off. People see it all the time.

That’s the oldest rule in direct response. People don’t buy information. They buy what it does for them. 193,355 hooks just put a number on it.

So what do you do with the stuff you actually want to teach?

You turn it into a discovery.

A lesson tells you how to do something. A discovery shows you something you didn’t know you were missing.

Same promise to save you time, told two ways:

One teaches. One feels like a secret you almost missed.

Both of these specific videos did fine. Any one post can get lucky. But post each kind a thousand times and the discovery wins more often.

Point that discovery at your health, open on a shocked reaction, and it jumps to 25.1% across 1,548 videos. Same combo as the “wait… you can bleed AND still be pregnant?!” hook I took apart earlier.

Reply with the word CATEGORY and what you sell. I’ll send back what’s breaking out in your corner. Free for the first few.

This was the biggest surprise.

Each of the three choices has a handful of options:

→ what’s in it for me? a discovery, a result, a joke, a story, feeling seen

→ why do I care? money, health, love, status, saving time

→ what do I see first? a number, a question, a reaction, a story already happening

One choice on its own can swing your odds anywhere from a few points to about 9. You’ll see the biggest swings below.

Stack all three and the spread gets huge. The best combo broke out 33.1% of the time. The worst: 9.1%. More than 3x.

The best one: a joke + status + a story already happening. Like this:

The worst one: nothing in it for me + saving time + no words on screen at all.

A few that cleared the bar:

Notice what’s missing. Nobody’s promising to save you time.

If you change nothing else, change these:

→ Make it about money or health, not saving time. 24.4% and 21.1% vs 15.4%.

→ Start with a number, not “how to.” 23.2% vs 17.1%. No words on screen at all: 15.3%.

→ Name a specific fear, not a dream. Fear: 23.6%. Hope and inspiration: 17.4%, across 24,242 videos.

I checked every video for the usual engagement tricks. Most of them barely matter.

→ “tag a friend”: 19.7% with it, 18.4% without

→ debate bait: 19.4% vs 18.6%

→ “pick a side”: 18.8% vs 18.5%

Two actually work:

→ leave in a small mistake people want to correct: 22.4% with one, 16.2% without

→ save the answer for part 2: 23.0% vs 18.5%

Both leave something open the viewer has to close. A mistake they can’t let slide. An answer they have to come back for.

Since the report came back, it’s shaped every ad concept we write. Some are already running.

Here’s what a rewrite looks like. One client writes about Gulf politics. Smart stuff. Their first hooks read like headlines.

One video covers a ruler’s public letter to his son, and what it really signals:

The first one tells you they have a clever way of reading the news. The second one drops you into a real story and makes you want the hidden part.

That rewrite hasn’t run yet. Ask me in a few weeks.

One client we rolled it out to: leads up 300% week over week, on the same spend.

This first pass read the words on screen and the caption. Not what people say out loud. Not what you see in the video.

And most of these 193,355 videos come from app, tech and consumer brand accounts. If that’s you, this is your data. If you sell something else, test before you trust it.

These are patterns to test, not laws. Test before you bet the budget.

I’m scraping and analyzing over 1,000,000 hooks next. What people say, what’s on screen, all of it.

That’s the part that gets me. For years a theory like puppet strings lived in my gut and a few client wins. Now Jev can read a video for a fraction of a cent, and I can check any idea against everything. The possibilities here are kind of endless.

That’s Report 02. More on that soon.

For now, Report 01 is free:

GET THE HOOK ATLAS REPORT (FREE)

12-slide report + a 3-page field guide. You get:

→ the hook taken apart, so the three choices click in 10 seconds

→ why the most common combo loses, with real examples

→ 6 hooks to copy, each one a real video that took off

→ the three swaps and the two tricks that work

→ a cheat sheet to check before you post

→ what works best in your category (9 categories)

→ a 60-second hook builder, five worked rewrites and a 10-minute testing sprint

GET THE HOOK ATLAS REPORT (FREE)

Go big,

Matt

P.S. Want my team to build your hooks and ad concepts off this? Work with us.