- The Visionary Group
- Posts
- Your AI ads aren't failing because of AI
Your AI ads aren't failing because of AI
The real problem isn't the model. It's what you hand it.
Hey, it's Patrick.
Think about this:
You open ChatGPT, type something like "write a 30-second UGC script for a health supplement targeting women 35–55," and get back a script that opens with: "Are you tired of feeling exhausted all the time? Introducing..."
It checks the surface boxes. Pain point. Product. CTA.
It also sounds exactly like every other ad your customer has seen for the last three years.
And in 2026, after years of being served AI-generated content, their nervous system has learned to tune it out before they even finish the hook.
The problem isn't using AI. The problem is what goes into the brief before AI ever enters the process.
Here's the framework we actually use and why the brief is the real leverage point.
Why the Brief Fails Before the Ad Does
Most creative briefs look like this: product name, key benefits, offer, tone of voice, maybe a reference video.
That brief goes to a creator found on a UGC marketplace. They've never used the product. They may not even be in the target demographic. They're going to perform the script as written because they're getting paid to… not because they believe it.
You can feel that. Your customer can feel that.
No research went into the brief. Nobody looked at what customers actually say about their pain in reviews, Reddit threads, or comment sections. Nobody checked the Meta Ad Library to understand which angles are already saturated in the market.
The brief is built on the brand's language, not the customer's language. And those two things are almost never the same.
The So What Test
Before we write a single word of creative at TVG, every brief has to answer one question:
If a cold customer sees the first two seconds of this ad, do they immediately know why it's relevant to their life?
Not interesting. Not clever. Relevant.
Most creative fails this test in the first frame. It opens with the brand name. Or a product shot. Or a hook that requires context the cold customer doesn't have.
If it doesn't pass the So What Test, the brief doesn't leave our desk.
The Research That Actually Feeds a Brief
Before the brief gets written, we do the work:
Meta Ad Library: Every competitor, last 90 days. What hooks are they running? What angles are saturated? Where's the white space nobody is owning?
Reddit threads, Amazon reviews, comment sections: We're looking for the exact words customers use to describe their pain. Not the brand's marketing language, the customer's language. Those are the words that stop the scroll, because they're the words already playing in the customer's head.
The Emotional Anchor (and the Part Almost No One Includes)
The brief is built around a specific emotional anchor, not around product benefits.
Two things go in that most brands never think to include:
1. The failed past attempts. What has this customer already tried that didn't work? When an ad names something the customer tried and failed at, something happens. They feel seen. They think: this person understands my situation. That moment of recognition is where trust begins.
2. Audience language, pulled verbatim. We lift phrases directly from the research.
Here's what that looks like in practice:
Generic brief | TVG brief |
|---|---|
Product: greens supplement. Benefits: energy, digestion, immunity. Tone: upbeat. | Pain anchor: Customer drags through the afternoon, on their third coffee, embarrassed their energy is worse at 38 than it was at 28. Failed attempts: sleep trackers, magnesium, cutting carbs. Hook angle: "I thought low energy was just part of getting older. Then someone on Reddit mentioned..." |
Same product. Completely different brief. The second one stops the scroll because the customer's inner monologue is the first thing they hear.
Where AI Actually Enters the Process
AI is a thinking partner here.
Once the research is done and the emotional anchor is locked, AI helps rapidly generate multiple hook angles built on that research. Not generic hooks. Hooks grounded in specific audience language, tested against the So What question.
Those hooks feed our five-module UGC script format: three hook variants at the top, a body that validates the customer's pain before introducing the solution, and a CTA connected to an emotional outcome.
Every module has a benchmark. Hook variants are measured against a 30% hook rate target. The body structure is designed to hold past the 50% mark. And every brief ships with a cold audience self-check before it goes to a creator.
AI didn't write those ads. But AI was embedded in the process that produced them as a research accelerator, a rapid variation engine, and a thinking partner that helps our strategists move faster without cutting corners.
The Real Framework
Do the research before you write the brief. Build the brief around the customer's language, not the brand's. Include the failed past attempts. Run every hook through the So What Test.
Then bring AI in to multiply what you already know.
The brief is where creative is either set up to win or set up to fail. Most brands have never seen what a properly built brief actually looks like.
If you want me to break down your account and show you exactly what's working (and what's holding you back), you can book a free audit here: Free C.O.R.E Growth Audit
We'll look at your brief process, your hook rates, and exactly where your creative is breaking down before it ever hits the feed.
Have a great rest of the week,
Patrick O'Driscoll
