I built an AI reading app for my 9-year-old son.
It wasn't supposed to teach me anything about AI visibility.
The idea was much more personal.
My son reads a book aloud to an AI agent. The agent listens. When he's finished, it asks him questions about what he just read and evaluates his comprehension.
At first glance, the interesting technology might seem to be the AI's ability to listen to him read.
But that's actually the easy part of the problem.
The important part comes next.
Did he understand what he read?
And to determine that, the AI needs to understand enough of what it heard to ask meaningful questions about it.
That distinction has been sitting with me.
Because the more I work on how businesses appear in AI answers, the more I think we're focused on the equivalent of hearing the child read.
We're celebrating access.
We're watching crawlers.
We're checking whether AI bots can reach our pages.
We're making content machine-readable.
All of that matters.
But there's a much bigger question:
After AI reads your website, does it actually understand your business?
Being crawled is not the same as being understood
This distinction matters because I've watched the conversation around AI visibility become increasingly technical.
Can the crawler access the page?
Is the content rendered?
Are AI bots blocked?
Is the information structured correctly?
Those are important questions. I pay attention to them too.
But a crawler reaching a page tells us something very specific:
It was able to reach the page.
It doesn't automatically tell us that an AI system understands the business.
It doesn't tell us whether that information will be retrieved when someone asks a relevant question.
It doesn't tell us whether the page will be used as a source.
And it certainly doesn't tell us whether the company will become a recommendation.
I've started thinking about these as five different stages. They are related, but they are not interchangeable — and they don't always happen in a neat linear sequence.
Crawled
Can an AI system access the information?
This is the technical foundation. If useful information can't be accessed, we've created a problem before the conversation even begins.
But access is only access.
Understood
Can AI make sense of the business?
What does the company actually do?
Who does it serve?
What is it particularly good at?
Where does it operate?
What evidence supports its expertise?
How do its services, people, customers, locations and areas of expertise relate to one another?
This is where I believe a lot of businesses have a bigger problem than they realize.
Their information exists.
But the meaning is fragmented across a homepage, an About page, service pages, old blog posts, PDFs, social profiles and language that assumes the reader already understands the company.
A human can often connect those dots.
We shouldn't assume an AI system will connect them the way we intended.
Retrieved
Now the context changes.
A buyer asks a question.
Not necessarily:
“Tell me about Company X.”
They might ask:
“Who can help my company understand why we’re disappearing from AI recommendations?”
Or:
“Which companies specialize in improving visibility in ChatGPT?”
Or dozens of increasingly specific questions as they move toward a decision.
Now the issue isn't simply whether AI knows a page exists.
Is the information relevant enough to this particular question to become part of the answer?
Cited
Then something even more interesting happens.
A specific page may be used as supporting evidence in an AI answer.
This is one of the things we pay close attention to in our work: which individual pages are actually being cited.
Not simply whether a brand was mentioned.
A mention tells me something.
A citation tells me something different.
The AI has selected a piece of content as a source for the answer it is constructing.
That doesn't mean the citation automatically produces website traffic, nor does it prove every stage before it happened in a neat linear sequence.
But it gives us a much more useful signal about which information is becoming useful inside AI-generated answers.
Recommended
And then we get to the stage businesses ultimately care about.
A buyer isn't asking for information anymore.
They're deciding.
Who should I hire?
Which company should I consider?
What's the best option for my situation?
Who should be on my shortlist?
Does your business become part of that answer?
That’s a very different standard from “Can ChatGPT access my website?”
This is why I don't think AI visibility is a content-volume problem
If we accept those distinctions, the content conversation changes.
The question stops being:
How much content should we publish?
And becomes:
What does AI need to understand about this business that it cannot understand clearly enough today?
That is a much more interesting problem.
Maybe the missing piece is a direct answer to an important buyer question.
Maybe the company's expertise is obvious to its existing customers but barely articulated on its website.
Maybe there isn't enough evidence behind a claim.
Maybe two important ideas exist on separate pages with no meaningful connection between them.
Maybe the founder has explained something brilliantly 50 times in customer conversations and not once on the website.
That last one became especially important to me.
Some of the best evidence is trapped inside people's heads
I've spent a lot of time talking to founders.
Ask a founder to describe their company in two paragraphs for a website and you'll often get polished marketing language.
Have a real conversation with them and something completely different happens.
They tell you why a customer actually chose them.
They explain the exception nobody put on the service page.
They tell you what buyers misunderstand.
They give examples.
They qualify their answers.
They explain why they disagree with the conventional wisdom in their industry.
They answer the questions customers ask over and over again.
That's expertise.
And much of it never becomes part of the company's public body of information.
It's one reason I've become so interested in preserving founder conversations and transcripts as source material.
A conversation can become an FAQ.
A buyer question.
A supporting article.
A deeper Learn page.
A comparison.
An explanation.
A piece of evidence that fills a gap in what AI can understand about the company.
The goal isn't to manufacture hundreds of pages from a transcript.
The goal is to surface knowledge that was already there but invisible.
The website has to tell a connected story
This is also why I've become less interested in looking at individual pages in isolation.
A business isn't one page.
Its expertise isn't one page.
A buyer's decision rarely comes down to one question.
The information has to work together.
The broad explanation establishes context.
Supporting pages answer more specific questions.
Founder and company information establish who is behind the expertise.
Examples and evidence support claims.
Comparison and decision-stage content help explain where the company fits and where it doesn't.
Each piece contributes something different.
The objective isn't to create a pile of content.
It's to create a body of evidence from which both humans and machines can understand the business.
And that brings me back to a child reading a book.
The question after the reading is the one that matters
When my son finishes reading to the AI agent, I'm not particularly impressed that the technology heard the words.
That's necessary.
But that's not the outcome I care about.
I want to know:
Did he understand what he read?
And the AI has its own version of that challenge.
Did it understand enough of what it heard to ask him the right question?
That's what made the connection so clear for me.
We have spent enormous energy making the web readable by machines.
The next challenge is making our expertise understandable enough to be useful.
For my son's reading app, the question is:
Did he understand what he read?
For businesses, I think the question is becoming:
Did AI understand what it read about us?
Because being crawled isn't the destination.
Neither is being cited.
The real business outcome happens when a buyer asks AI a question that matters — and your company has provided enough clarity, context and evidence to deserve a place in the answer.
After AI reads your website, does it understand your business well enough to know when you should be part of the answer?
That's the question I'm building around.