AI

I Tested 5 Hooks on One Piece of AI Content. Here Is the Winner.

October 30, 2026 · Brian Arfi Faridhi

I had one piece of content. Then I wrote five different hooks for that exact same content.

This is not a flex. I wanted to know which one makes people stop scrolling, and which one makes them keep going. More importantly, I wanted my team at the office to learn how to measure instead of guess.

Here is the full experiment, opened up for you.

Why test hooks only

Good content with a weak hook is a waste. People never reach the body, because they are gone in the first two seconds. On the flip side, an honest but punchy hook can carry ordinary content all the way to the end.

So I did the thing I always ask from product teams. One variable, several variations, compare the results. The content stayed identical. Only the opening line changed. Five versions.

Producing them was fast. I sat down with AI, gave it the context of the content, and asked for hook candidates. I did the picking and the polishing. Work that used to eat half a day of solo thinking was done in an hour. The pattern holds everywhere: what used to take 2 days now takes 2 hours.

The rules, so you can copy this

If you want your team to run this, here are the rules I used.

First, every hook must be specific. At least one real number or name. "AI tips for work" is not a hook, it is a category. "The PM with 3 years of experience who manually writes updates every Monday" is a hook.

Second, the hook must call out who it is for. Fresh grads, programmers afraid of being replaced by AI, people selling online. The reader should think, this is me.

Third, and this is the one I use most, the 10X test. I write a normal hook first, then ask AI: "What is the 10x more extreme version?" What gets amplified is the stakes, meaning the worst consequence if the reader does nothing. The event itself stays undramatized. Example: "you are probably taking too long to write reports" becomes "every week you are paying your own salary just to write reports AI could finish in minutes."

Extreme versions that stay honest, I use. Versions that become exaggerated, I throw away and go back to the normal one.

What I learned from the 5 versions

Three lessons came out of this, and all of them apply if you want your team using AI.

Lesson one: hooks that name a real work situation beat hooks that stay vague. "5 hooks, 1 piece of content" lands better than "the secret to viral content". Concrete numbers make a testable promise. Jargon does not.

Lesson two: AI generates variations fast, but the human taste still has to be mine. Out of dozens of AI candidates, maybe three were actually usable. The rest sounded like ads. If you hand the entire decision to AI, everything comes out average and nothing stands out.

Lesson three: this kind of iterate-and-test loop is the easiest form of AI work for a team to accept. Nobody needs to be an engineer to join. PM, marketing, ops, anyone can run the same format. Make 5 variations, test, see which one wins. It is a softer entry point than telling everyone to "go learn AI".

How this connects to $4 million per year

Some of you will ask: what does a small experiment like this have to do with the big number on my site?

Saving companies money is an old habit of mine, long before AI was trendy. The total is $4M+ per year, from a mix of cost-saving initiatives I drove as a product leader: process improvements, cost optimization, more efficient product decisions. Automation is only one of them, not the whole story.

Back then, this kind of leverage required a senior position, an engineering team, and expensive systems. I had to become a product leader first before I could push change like that. Now AI is the sharpest tool for the same habit, and it can be trained into everyone on the team. The 5-hook experiment is the smallest possible version of it. The pattern is identical: measure, compare, keep the winner, repeat.

My current proof that this works in the AI era: an 8-channel content distribution engine I built and run by myself, my Applied-AI Certification, and AI Circle. Same principle, new tooling, no engineers required.

Where to start

If you want your team using AI, do not start with theory training. Start with one small experiment that shows results within a week.

The format is simple. Pick one repeated task, like writing hooks, captions, or meeting summaries. Make 5 variations with AI. Test on the channels you already have. Note which one wins. Discuss with the team in 30 minutes. Repeat next week.

Within a month, your team has its own data, not just motivational stories about AI.

If you want to practice patterns like this with me and a community running the same experiments, check out AI Circle. If you are a decision maker and want your team to learn this AI experimentation method directly inside your company, see the corporate training page.