The question most business owners have never asked. The answer is shaping which of them get customers and which of them do not.
Ask ChatGPT what it knows about a small business in your neighborhood. Then ask Perplexity. Then Gemini. Read what each one says. Check it against what you actually know about the business. Note what is right, what is wrong, and what is missing.
Almost no owner of that business has done this. The ones who have, often by accident, tend to describe the experience the same way. Some version of unsettled.
“I thought we were doing fine,” said Melissa Tanner, who runs a family dental practice in the Dallas-Fort Worth area and tested her own visibility last fall after a patient mentioned finding a competitor through ChatGPT. “We were ranked on Google, we had reviews, we had a website that had been updated that year. And then I asked ChatGPT who the best family dentists in our area were, and we weren’t in the answer. Not in one of them. Not in any of them.”
The gap Tanner describes is becoming more common. It is also becoming more expensive.
A new discovery channel that most owners have not tested
Consumer research conducted in 2026 by BrightLocal found that 45 percent of consumers now use artificial intelligence tools to find local services, a number that stood at 6 percent a year earlier. Pew Research has documented that roughly 39 percent of U.S. adults now use ChatGPT weekly for decision-making tasks that include comparing products and services.
The businesses being named in those AI responses are capturing an increasing share of high-intent customer inquiries. The businesses not being named are, in many cases, unaware that the inquiries are happening at all.
“There’s no bounce in your analytics when a customer asks ChatGPT for a recommendation and doesn’t get your name,” said Rand Fishkin, co-founder of the analytics firm SparkToro, whose research team published one of the most comprehensive studies to date on how AI platforms generate brand recommendations. “The customer never arrives. The loss doesn’t register anywhere.”
What Fishkin’s January 2026 research established, drawing on nearly 3,000 test prompts run across ChatGPT, Claude, and Google’s AI, was that AI recommendations are probabilistic rather than ranked. A business appears in the list of recommended options when the AI has enough confidence about it across the sources the model consulted. A business without that confidence sits outside the list, often for reasons the owner has no way of diagnosing without testing directly.
The ten-minute test that produces clarity
The test itself is simple, and it yields three distinct pieces of information most owners have never seen before.
The first is whether AI platforms recognize the business exists. An owner can open ChatGPT and type a prompt asking about the business by name. Something along the lines of “Tell me about [business name] in [city].” The response falls into one of three categories. The AI produces no recognition, either stating it has no information or generating a generic answer that does not describe the actual business. The AI produces partial recognition with errors, describing the business but including incorrect details about location, services, hours, or ownership. Or the AI produces accurate recognition, describing the business correctly.
The second is whether the AI recommends the business when asked about its category, without being prompted by name. An owner asks the AI the question a prospective customer would ask. “Who are the best family dentists in Fort Worth?” “Recommend a personal injury attorney in Nashville.” “I need an HVAC company in San Diego that handles commercial work.” Running each prompt three to four times, because AI responses vary, reveals whether the business shows up consistently, occasionally, or never.
The third is whether the information the AI has, when it does mention the business, is accurate. Prompts asking for specific details, services offered, hours, location, clientele served, produce answers that can be checked against reality.
“The reason you have to test all three,” Tanner said, reflecting on her own experience, “is that we passed the first one but failed the second. ChatGPT knew we existed. It just never recommended us when someone asked for a dentist. And that’s where the lost patients were.”
The gap between existing and being recommended
SOCi, a marketing technology firm that analyzed more than 350,000 business locations across 2,751 brands in its 2026 Local Visibility Index, found that ChatGPT currently recommends roughly 1.2 percent of local businesses when asked category-level questions. Google Gemini, which draws more heavily on Google Maps data, reaches about 11 percent. Perplexity sits at 7.4 percent. Google’s traditional local search results, by comparison, give some form of visibility to roughly 35.9 percent of businesses in each category.
The gap has practical consequences. Monica Ho, chief marketing officer at SOCi, has framed the stakes bluntly in public statements. “Consumers aren’t scrolling through options anymore. They’re asking AI to decide for them. If your brand isn’t optimized for AI search, you’re not just losing rank, you’re removed from consideration entirely at the moment of intent.”
The businesses inside that 1.2 percent on ChatGPT share certain traits, according to SOCi’s research. Consistent business information across directories, review platforms, and their own websites. Recent review activity distributed across multiple platforms rather than concentrated on Google alone. Content that directly answers the kinds of questions prospective customers ask AI assistants. Structured data that tells AI platforms what the business is and who it serves.
Most small businesses have some of these signals and not others. The test reveals which.
Two categories of companies responding to the shift
The market has responded to the AI visibility gap with two distinct categories of companies, and industry observers note that the difference between them is becoming more visible to business owners trying to navigate the space.
The first category is monitoring. Companies like Profound, AthenaHQ, and AI Rank Checker provide dashboards that track how often a brand appears across AI platforms, what competitors are being named instead, and how visibility shifts over time. These tools have grown rapidly as owners become aware of the gap and want a way to measure it.
The second category is execution. A smaller set of firms, including companies like Yext on the data infrastructure side and Yazeo on the strategic execution side, argue that monitoring the gap without closing it leaves businesses no better off than before they knew it existed. Yazeo has built its positioning around a direct challenge to the dashboard-first model, arguing that businesses serious about being recommended need the underlying citation, content, and review work done rather than simply observed.
“The reports are accurate, but they don’t build anything,” said a spokesperson for one execution-focused firm in the space, describing the distinction between the two categories. The argument, increasingly common among execution firms, is that the AI visibility market is moving toward a split between diagnosis and treatment, with most businesses eventually needing both.
The distinction matters to owners like Tanner. “I didn’t want another report telling me we had a problem,” she said. “I wanted the work done.”
What the test does not tell you
The ten-minute diagnostic reveals where a business currently stands. It does not tell an owner how to close the gap, and it does not produce instant results.
Closing AI visibility gaps typically takes 90 to 120 days of sustained work, according to multiple firms operating in the space, with meaningful improvement visible in most cases within six months. The work involves cross-referencing business information across directories, restructuring website content around the kinds of questions prospects actually ask, building citation depth through independent sources, and diversifying review activity across platforms. None of it is fast, and some of it requires specialized knowledge most owners do not have in-house.
But the test itself, running those three prompts across four AI platforms and recording the answers, is the beginning of most successful closures. Owners who have never looked cannot act. Owners who look, according to those who have done it, tend to move faster than they otherwise would have.
“It took ten minutes,” Tanner said. “I wish I had done it a year earlier.”
The test takes ten minutes. The shift in local business discovery that makes the test worth running has been building for the past eighteen months, and it shows no signs of slowing. Consumer research suggests the 45 percent of Americans currently using AI to find local services will continue to grow, with Gartner projecting that roughly a quarter of all organic search traffic will shift to AI chatbots and voice assistants by the end of 2026.
Whether a business is inside the answer those platforms give, or outside it, is one of the more consequential questions an owner can ask in this environment. Most have not asked. The ten minutes it takes to find out may be the most informative ten minutes of the quarter.