AI can now make 50 versions of a still ad in one minute.
Someone still has to decide which one gets the media money.
Fifty Dishes Don't Make a Good Plate
We saw this in our own study. Per Kasch of SightEffect made our test pictures with AI, ten versions per brand, identical except for one thing. Producing versions was the easy part. Choosing between them still runs the way it did ten years ago: a meeting, a liking score, and in the end the most senior opinion in the room. In our webinar poll, about half the room chose by discussion and gut feel.
It's like a buffet. Fifty dishes don't make it easier to put together a good plate. They make it easier to fill one.
So the scarce thing is no longer the picture. It's the rule for picking it. Fifty versions need one good rule.
"Attention-Grabbing" Didn't Drive Purchase
"Attention-grabbing" is in almost every picture brief. Open yours and check.
In our test for AXA and ChatGPT it didn't drive purchase. Per and I measured which picture qualities do, for an AXA pension product and the ChatGPT subscription, with 1,000 people in Germany. It was an independent study, not commissioned by AXA or OpenAI. We measured gut reactions, which is what people feel before they start explaining. Then Causal AI worked out which of those reactions actually drive the wish to buy.
If you read our earlier essay on look and feel, this will sound familiar: the space-suit child scored high on interesting and low on honest. Interesting is not worthless. It just isn't the whole job.
The 2+2 Rule
Four qualities came out of the data. Two worked in both categories:
- Interesting. The picture earns a second of relevant attention. Relevant is the operative word. Attention for its own sake is entertainment.
- Trustworthy. Every purchase carries a risk, and a trustworthy picture lowers it before anyone reads a word of copy.
The other two follow the buyer's worry, and that one is different for every category:
- AXA (pension): honest and secure. The worry behind it is probably: "Will my money be safe in 30 years?"
- ChatGPT (subscription): competent, and fits the brand. The worry behind it is probably: "Is this the real thing, and is it good?"
It works like choosing a doctor and a plumber. You need to trust both. After that you want the doctor careful and the plumber on time.
We call this the 2+2 Rule. Two qualities that work everywhere, two that your buyer's worry decides.
Two Categories Are No Law
I won't sell this as a law. Two categories don't make one, and these are reported results from one study, not audited facts.
What I will say is the hypothesis. Your second pair comes from your buyer's main worry. A pension buyer fears the 30-year horizon. A buyer of an AI subscription fears paying for something that doesn't deliver. Same kind of funnel, different reasons to buy.
The 2+2 Rule as a creative-review filter
- First 2, always: Is it interesting for the right reason? Does it feel trustworthy before anyone reads the copy?
- Second 2, per category: What is the buyer's main worry, and which two qualities answer it?
- Test the effect: Does the version move purchase, or just liking?
You Can't Ask for the Worry
Here is the catch. You won't get that second pair by asking people directly. "Why do you buy?" produces the reasonable answer, and the reasonable answer is rarely the one that moves them. That's the say-do gap, and it is why customers can't tell you why they buy.
What drives a purchase is implicit: subconscious goals on different levels. You find them in an indirect way. That's what deep implicit research does. In this study it measured reaction time on 1,000 respondents and fed the result into a Causal AI model.
For AXA's retirement product, the answer came out as trust and honesty. For ChatGPT, competence. Two categories, two completely different reasons to buy.
That homework is necessary before anyone opens an image tool. Skip it and your creative is decorating a guess, however good the picture looks.
Five Questions Before the Next Picture Goes Live
If you take one thing from this essay, take this list into your next brief:
- Do we know what our current key visual does to purchase, as a number?
- Is the picture interesting enough to earn a second of real attention?
- Does it feel trustworthy before anyone reads the copy?
- What is our buyer's main worry, and which two qualities answer it?
- Do we choose between variants by liking or by effect on purchase?
Question 4 does most of the work.
Someone asked in our Q&A whether that means testing every picture. It doesn't. Do the homework once and find what drives purchase in your category. After that, new pictures can be checked against those few qualities, with small tests or with a customer twin trained on that study.
The Same Homework Sits Under the GTM Quote
A line went around recently: "Branding is how people see you. Marketing is how they find you. Sales is how you get paid." It collected around 20k likes and 7k saves, by the numbers shown on the post. I think it's the modern but wrong understanding of go-to-market.
Look at what each line covers. Branding becomes a result, with no word on where it comes from. Marketing shrinks to promotion. Sales becomes the point of sale. Real marketing is the four Ps, product, price, place and promotion, and it starts with a strategy that builds on solid customer understanding.
All of it rests on something the quote leaves out: knowing why your customers buy. Without that, branding builds an image nobody was looking for, and promotion brings in people the product can't satisfy. The 2+2 Rule is that insight, applied to one picture.
What to Do Monday Morning
Write your category's answer in one sentence. If your 2027 brief can't name the buyer's main worry in a single line, the brief isn't finished.
Then ask the question that matters more than any A/B result: how did you find out? If the answer is "we asked," or "we felt it," you have a guess. If you want to see where your own creative decisions rest on stated opinions rather than measured drivers, the AI Diagnostic is the quickest start. For the thinking behind it, see why awareness is a vanity metric and my book THE TOP 5%.
Liking or effect on purchase: how does your team pick today?
The 2+2 Rule: frequently asked questions
What is the 2+2 Rule for ad pictures?
The 2+2 Rule is a hypothesis from SUPRA and SightEffect's picture research. Two qualities worked in both categories tested: the picture is interesting (it earns a second of relevant attention) and trustworthy (it lowers purchase risk before the copy is read). Two more follow the buyer's main worry and differ by category. Dr. Frank Buckler presents it as a hypothesis from two categories, not a law.
Which picture qualities drove purchase for AXA and ChatGPT?
In the study of 1,000 people in Germany, interesting and trustworthy worked in both categories. For the AXA pension product, the two category-specific qualities were honest and secure. For the ChatGPT subscription, they were competent and fits the brand. The pattern points to the buyer's underlying worry: money safe in 30 years for a pension, a real and good product for an AI tool.
Why does attention-grabbing not drive purchase?
In SUPRA's test for AXA and ChatGPT, attention-grabbing as a quality did not drive purchase. What mattered was relevant attention: a picture interesting for a reason aligned with what the product does for the customer. Attention on its own is free entertainment. Trust and the qualities that answer the buyer's worry turned that attention into a wish to buy.
How do you find your category's buyer worry without asking customers?
Direct questions return the reasonable answer, which is rarely what moves people. The study measured gut reactions with implicit reaction-time tests on 1,000 respondents and used Causal AI to separate the perceptions that drive purchase from those that merely correlate. Do this homework once per category. Later pictures can be checked against the few qualities found, with small tests or a customer twin.
Dr. Frank Buckler advises leadership teams on growth decisions that depend on why customers choose. He founded SUPRA, holds a PhD in marketing and was among the first to apply causal AI to marketing questions. His latest book is THE TOP 5%.