An AI mention is only the beginning

When I first started working in AI SEO, a lot of the conversation was about getting your business mentioned in an AI answer. People were testing prompts, tracking which brands appeared, and trying to figure out how to show up more often. That’s still a big part of the conversation.

I was interested in that too. But I kept wondering what happened after the mention. If someone asked a few follow-up questions, could the AI explain why that business was a good fit? Did it actually have the right information to support the recommendation?

Because that’s where I think this gets interesting.

Someone asking for a recommendation might still be figuring out what they want. They ask a question, read the answer, add some context, and ask another question. The business that made sense at the beginning might make less sense once the person explains their situation.

I want to understand what happens throughout that conversation.

What happens when the customer gets specific?

Take real estate, since it’s a space I know.

Someone asks AI to recommend an agent in their area. Your name comes up. That’s worth knowing.

Then they say they’re selling a waterfront property. They want someone who understands that market, has relevant experience, and can explain how they’d approach the sale.

Now there’s more to work with.

Can the AI give specific, accurate reasons you might be a good fit? Can it distinguish your experience from the other agents it mentioned? What happens when the person asks, “Why would I choose this agent over that one?”

I’d learn a lot more from those answers than from seeing a name in the first response.

The recommendation might hold up. It might turn out to be based on something vague. It might include information that sounds convincing but isn’t correct. Or the customer might introduce a requirement that makes a different agent the better choice.

All of those outcomes tell us something useful, provided we actually look at them.

Does AI have enough information to understand the business?

This is one of the first things I’d want to investigate.

A business can have a real strength that barely exists in its public information. Maybe it has experience with a particular type of customer. Maybe there’s a service it handles especially well, or a limitation that matters when deciding whether it’s a good fit.

But its website says the same broad things every other website says.

“Great service.” “Local expertise.” “Personalized solutions.”

Okay, but what does that mean for the person asking?

If I have to talk to the owner for twenty minutes to understand why someone should choose the business, there may be useful information missing from the website and other public sources. I’d want to identify that gap before assuming we need some clever AI SEO tactic.

What does the business actually do? Who does it serve well? What makes it different in a way a customer would care about? And what evidence do we have for those claims?

Those questions give the work some direction.

They also help separate a genuine advantage from something we just like saying about ourselves.

I care about the reason behind the recommendation

I want AI to understand a business accurately enough to recommend it when it makes sense.

That includes finding the specific strengths that matter to the right audience. There can be a lot of value in explaining something clearly that the business has struggled to communicate.

But I don’t want to force a recommendation where the fit isn’t there.

If someone needs a service you don’t offer, the answer should reflect that. If another business is better suited to their situation, I’d expect the recommendation to change.

That doesn’t make the work unsuccessful. It makes the information more useful.

Where I see the opportunity is when a business really is a strong fit, but the information available doesn’t make that clear. The system might know the name without having enough context to explain why it belongs in that particular conversation.

That’s a much more specific problem to work on than “we need more mentions.”

What would convince me it’s working?

A screenshot can be a useful observation. I just wouldn’t ask it to prove more than it does.

I’d want to know what the person asked, whether the business was already named in the prompt, and whether the answer got the facts right. Then I’d want to see what happened as the questions changed.

Did the AI support its recommendation with accurate reasons? Did it recognize a limitation? Did it keep repeating a claim after the customer introduced information that should have changed the answer?

I’d want to look across realistic customer situations and repeat the observations, with a record of the questions, dates, systems, and how we evaluated the answers. Otherwise, it’s too easy to find one response we like and build an explanation around it.

I’m comfortable having a point of view before every part of it is proven. This is mine: evaluating the conversation gives us a better way to investigate whether AI understands a business than counting an initial mention alone.

What I wouldn’t do is assume that better answers automatically mean more customers. That’s another thing we’d need to measure.

Start by asking another question

If I were looking at a business today, I’d start with the questions its customers actually ask.

I’d make the situation specific. Give the customer a budget, a location, a particular need. Ask why the business fits. Ask where it might not.

Then I’d check the answers against the business itself and the information available about it.

You might find an error worth correcting. You might find that an important strength is poorly explained. You might realize your own website leaves people with the same unanswered questions.

None of that guarantees a particular AI response. It does give you something concrete to investigate and improve.

I still care whether a business gets mentioned. I just don’t think that’s where the evaluation should end.

If someone asks, “Why them?”, I want there to be a good answer.