Morgan Housel has a very uncomfortable idea in The Psychology of Money.
We know a tiny fraction of what is happening.
Something like 0.0008%.
But our brains behave as if we have 98% of the picture.
Which is bold.
Also, a useful explanation for most marketing meetings.
Humans Hate Incomplete Information
So we build stories.
Quickly.
Confidently.
With dashboards.
With “industry benchmarks”.
With what worked in 2019.
With one person saying, “our brand is strong”, in a tone that should probably require a traffic audit.
We see this all the time.
By the time anyone asks “how visible are we to AI?”, the business already has a story.
We rank on page one.
Our SEO agency has it covered.
ChatGPT obviously knows who we are.
AI traffic is tiny anyway.
It’s just the new SEO.
We’re fine.
And sometimes the story is right.
But sometimes it is just a beautiful house built on four facts and a mood.
Here’s the Twist: The AI Has a Story Too
While your team was building its story about your visibility, the models were building their story about your business.
From fragments.
A crawled product page here. A Reddit thread there. A review from three years ago. A competitor’s comparison table that was, let’s say, generous to the competitor.
Ask ChatGPT what your company does.
Not what you hope it says. What it actually says.
Ask it what you sell. Who you’re for. What you cost. Whether you ship to Germany. Whether you’re better than the brand it names next — because it will name one.
For most merchants, this is the first uncomfortable meeting of the two stories: the one you tell about yourself, and the one an AI tells 700 million weekly users on your behalf.
They rarely match.
The model’s story might be outdated. Incomplete. Confused with a similarly named company. Or — most common — simply missing. You asked about the best options in your category, and you were not in the answer.
You didn’t get a rejection notice. You just weren’t in the story.
The Job Is Not to Kill the Story
That’s too easy.
Pessimism about AI channels is as lazy as blind optimism. “It’s all hype” is also a story — built on the same four facts and a slightly different mood.
The job is to test whether the story survives contact with reality.
What if the model describes your business wrong?
What if it recommends your competitor by name?
What if your “page one ranking” no longer produces clicks because 60-70% of searches end without one?
What if an AI shopping agent tries to buy from your store and gets stuck at step two?
What if the upside of AI-referred traffic is real — it converts at 4.4x organic — but nobody priced the downside of being invisible while it compounds?
- 700M weekly ChatGPT users hear the model’s story about your brand
- 60-70% of searches now end without a click
- 4.4x organic conversion — the rate of AI-referred traffic
These are not rhetorical questions. Every one of them is checkable this afternoon, for free, by typing your category into ChatGPT and reading what comes back.
Almost nobody does it.
Because the story feels like knowledge, and checking feels like doubt.
Falling Out of Love with the Wrong Story
The most useful thing anyone in your organization can do right now is not to say “no” to AI channels, and not to say “yes”.
It is to help the business fall out of love with the wrong story before it becomes an expensive one.
The mechanics are boring on purpose:
Ask the models. ChatGPT, Gemini, Perplexity, Copilot. Ask what your business does, what you sell, and who they’d recommend in your category. Write down what they get wrong. That delta is your brand’s AI visibility gap, in plain text.
Check the funnel you can’t see. AI-referred visits hide in your analytics under direct and dark traffic. Impression-to-click ratios in Search Console tell you how much of your “visibility” has quietly stopped producing visitors.
Test the transaction, not just the mention. Being described correctly is table stakes. The channel that matters is the one where agents complete purchases — can an agent find your products, verify inventory, and check out? If not, you’re a citation, not a store.
Fix the source material. Models tell the story your structured data lets them tell. Comprehensive, consistent, machine-readable product data is how you edit the story the AI tells about you — it’s the one input you fully control.
The Most Dangerous Sentence
Because the most dangerous sentence in brand strategy is not:
“We don’t know how AI sees us.”
That sentence starts audits. It gets things checked. It’s the beginning of every visibility gap that ever got closed.
The dangerous one is:
“We know enough.”
That’s the sentence that was said about mobile in 2010, about marketplaces in 2015, and about zero-click in 2023 — always in a confident tone, always with a chart, always about 18 months before it got expensive.
You know 0.0008% of what’s happening.
So does everyone else.
The advantage doesn’t go to whoever has the best story.
It goes to whoever tests theirs first.
Want to see the story AI tells about your store — including where agents get stuck trying to buy? Run the Agent Commerce Simulation, or talk to us about closing the gap between your story and theirs.
