An AI Citation Doesn't Pay the Bills

A report lands in an owner's inbox: "Your business was mentioned by ChatGPT 14 times this month, up from 6." It reads like a scoreboard going the right direction. Nobody attached a number for how many of those 14 mentions turned into a phone call. Nobody asked.

Here's the thing nobody selling Answer Engine Optimization (AEO) or GEO wants to say plainly: a mention is not a metric you can pay rent with. It's evidence something is happening. It is not, by itself, evidence of anything an owner should be writing a check for.

What "AI visibility" actually measures

"AI visibility" — the term vendors, agencies, and dashboards use for how often ChatGPT, Perplexity, Google AI Overviews, or Gemini name a business — measures exactly one thing: how often your name showed up in an answer. That's a real, trackable event. It is also several steps removed from a customer.

Being cited and being recommended are already two different outcomes decided by two different steps of the same system — a business can get pulled in as a source and still lose the actual recommendation to a competitor. A citation count doesn't tell you which of those happened. It just tells you the business got mentioned somewhere in the answer, favorably or not, chosen or not.

That gap is exactly what Kevin Lee's column was pointing at: an industry that's gotten very good at reporting the mention and has mostly stopped asking whether the mention did anything. A dashboard showing citation counts climbing month over month looks like progress. Whether it correlates with a single additional phone call is a different question the dashboard was never built to answer.

Why this matters if you're the one paying for it

If you're a local service business — a plumber, a groomer, a dentist, a contractor — the thing that actually pays the bills was never "getting mentioned." It's the call that gets booked, the form that gets filled out, the walk-in who says "I saw you were the one people recommended." Visibility is upstream of that. It isn't a substitute for it.

I ran a dog grooming salon for years before any of this existed, and I can tell you exactly how much I cared whether people were "aware" of the salon in the abstract: not at all. I cared whether the appointment book was full next Tuesday. Awareness that doesn't turn into an appointment is a nice feeling and a slow way to go out of business. AI visibility, without a path to an actual booked job, is the same problem with a newer name.

This is where a lot of AEO and GEO advice — and a lot of what gets sold under those labels — quietly goes sideways. It's genuinely easier to report a citation count than a booked-job count. Citation counts are visible in a tool, they climb reliably with the right content work, and they make for a clean chart in a monthly report. Booked jobs require actually asking the customer how they found you, which is more annoying and less automatable, so it gets skipped. The metric that's easiest to report becomes the metric that gets reported, whether or not it's the one that matters.

The mechanics: what to actually check

None of this means ignore AI visibility — getting cited is still the entry ticket, and a business that never gets named by these tools has a bigger problem than measurement. The fix is checking visibility and outcomes as two separate questions, not one.

1. Decide what "working" means before you look at any dashboard. For most local businesses that's one of three things: a phone call, a form submission, or a walk-in who mentions they looked the business up first. Pick the one that's real for your business before you go looking for proof.
2. Ask, don't infer. "How'd you hear about us?" is a low-tech question that still works better than any analytics workaround for AI-referred traffic. If someone says "I asked ChatGPT for a recommendation near me," write it down. A running list beats a guess.
3. Watch it separately from the citation count, not instead of it. A citation count going up and a call count staying flat is useful information — it says the visibility work is landing somewhere that isn't converting yet, which is different from the visibility work not landing at all.
4. Ask anyone selling you AEO or GEO work what outcome their reporting ties to. If the entire monthly report is mentions, citations, or a proprietary "visibility score" with no line connecting it to a call or a booking, that's worth asking about directly — not necessarily a dealbreaker, but a gap worth naming out loud.
5. Give it more than a month. Chasing attribution on a single month of AI-referred traffic is its own trap — that's a different, harder problem than the one here, which is simpler: are you even tracking the outcome that matters, on any timeline, or only the mention.

What won't help

• Treating a citation count or a vendor's "AI visibility score" as the deliverable. It's an input, not an outcome, regardless of how polished the number looks in a monthly report.
• Firing everything AI-search-related the first month a citation count doesn't turn into a call. Visibility work compounds; one flat month next to a rising citation count is a reason to keep watching, not a verdict.
• Assuming more mentions automatically means more business. It's a reasonable-sounding assumption. It's also the exact assumption Lee's column was written to push back on.
• Skipping visibility work entirely because it's "just marketing metrics." The opposite overcorrection is just as broken — a business that's never mentioned by these tools has nothing downstream to measure in the first place.

FAQ

Is AI visibility worthless if it isn't producing calls yet?

No. It's a necessary but not sufficient condition — a business has to be mentioned before it can be chosen. The mistake is stopping the measurement at "mentioned" instead of following it to "chosen."

What should I ask an agency or vendor selling AEO or GEO work?

Ask what outcome their reporting connects to beyond citation counts, and ask how they'd know if the work stopped producing business even while the visibility numbers kept climbing. A vendor with a real answer has thought about this already.

How is this different from figuring out if my AEO spend is working?

That's a related but separate question about measurement mechanics — your analytics genuinely can't see most AI-referred traffic. This is a step earlier: deciding what you're even trying to detect before you go looking for it, so you're not celebrating a metric that was never going to pay the bills regardless of how well you measure it.

What's a realistic sign that AI visibility is actually producing business?

A caller or customer who mentions asking an AI tool for a recommendation, a "how'd you hear about us" answer that names ChatGPT or a similar tool, or a Search Console jump in people searching your business name directly after an AI mention. None of these are perfect. All of them are better than a citation count alone.

Should I stop tracking AI visibility altogether and just count calls?

No — visibility is still the input. Tracking it tells you whether the work is landing anywhere at all. The point isn't to stop watching it, it's to stop treating it as the finish line.

Does this apply if my business books online instead of by phone?

Yes, the mechanics just shift. Swap "phone call" for "booking form submission" or "checkout started," and ask the same question: is there a path from an AI mention to that specific action, or just a mention sitting on its own.

How often should I check whether visibility is turning into business?

Monthly is reasonable for most local businesses — frequent enough to catch a real shift, infrequent enough that you're not overreacting to normal week-to-week noise in either number.

---

This is the part of Answer Engine Optimization that's easy to sell short: getting a business named by an AI tool is genuinely useful work, and it's also not the same job as making sure that mention turns into something that pays. Getting found in Google and AI search covers both halves — the visibility work and the follow-through — instead of treating a rising citation count as proof of anything on its own.