by 
Original AI Search Research · What We Learned
What We Learned About AI Search by Analyzing 28,488 Businesses
When someone asks ChatGPT, Claude, Gemini or Google which businesses to consider, what actually determines who appears? We analyzed 28,488 businesses to see how AI search behaves in the real world, why the answers change and what businesses can do to improve their chances of being discovered.
AI search does not produce one fixed ranking. The businesses that appear can change with the question, platform, location and context. What matters is whether your company keeps showing up when buyers ask the questions that matter to them.
The research at a glance
One AI search result is not enough to understand visibility
One ChatGPT result can be encouraging, but it only tells you what happened in that one search. We wanted to see what held up when we changed the question, industry, city and AI platform.
Why we did it
We needed a better way to understand how AI search behaves
AI search is now one of the ways people discover and evaluate businesses, but we kept seeing the same kind of proof offered over and over: someone asked ChatGPT a question, saw their company in the answer, and treated that one result as proof they were “winning AI search.”
We did not think that told us enough.
Traditional search trained marketers to think in rankings. You can look at a query, see a position and track how it moves over time. AI recommendations are much more fluid. A business can appear for one question and disappear when the wording changes. It can show up in ChatGPT and not Claude. It can appear for a broad recommendation and not for a specific need. In Google, the AI Overview may not appear at all.
We focused on two things: which businesses kept showing up, and what made the list change?
We tracked repeat appearances, platform differences, question wording, location and industry.
How the study worked
The methodology got messy quickly
We could have asked one platform one question, recorded the names and published a leaderboard. That would have missed most of what makes AI search hard to measure.
Build the business list
We started with a defined list of businesses and applied the same industry and active-business rules across the study.
Hold geography steady
We used nine exact Florida cities rather than metro areas, counties or radius estimates so we were comparing the same kind of market each time.
Change the question
Each city and industry was tested three ways: a broad “best” search, a direct recommendation request and a more specific need-based question.
Change the platform
We ran the searches across ChatGPT, Claude, Gemini and Google AI Overviews, the AI-generated summaries that sometimes appear in Google search. One platform cannot represent AI search as a whole.
Match recommendations to real businesses
AI answers do not always use the exact business name or location in our list, so every recommendation had to be matched carefully before we counted it.
Expect the results to move
AI systems change over time, so we treated the study as a large snapshot of recommendation behavior. We expect the names and results to move.
The final study included 28,488 businesses. We counted a business only when we could clearly match an AI recommendation to the business in our study list. Local businesses missing from the starting list and businesses outside the exact city were tracked separately.
We also found relevant local businesses in the AI answers that were not in our starting database. No business list is complete, and AI systems are not limited to the database we happened to start with.
What the data showed
Why AI search results change from one question to another
Only 5.9% of the businesses in our study appeared at least once. Even within that small group, visibility changed a lot when we changed the question or the platform.
appeared at least once
1,676 of the 28,488 businesses appeared in at least one tested recommendation. The other businesses did not appear in the searches we tested.
appeared in exactly one search
Among businesses that surfaced, a single appearance was much more common than repeated presence across the 12 searches available for their city and industry.
appeared on only one platform
Most visible businesses showed up on only one AI platform. Only 5.7% showed up across all four.
were tied to one question type
When a business appeared on a platform, it usually showed up for only one of the three question types.
How often did visible businesses keep showing up?
How many of the 12 available searches did a visible business appear in?
Showing up once is good. Showing up repeatedly is what matters.
A more specific question can produce a different list
For every city and industry, we tested three kinds of buyer questions: a broad “best” search, a direct request for recommendations and a narrower question built around a specific need. That last version often produced a genuinely different set of businesses. Among the local businesses that appeared for the specific-need search, 37.4% had not appeared in either of the two broader searches.
Think about the difference between asking for the “best CPA firm in Tampa” and asking for a “CPA firm that works with manufacturers operating in several states.” Those are different problems, so it makes sense that the recommended businesses may change. A firm that never appears on the broad list could become much more relevant once the buyer adds an industry, service, location or other specific need. That is why testing one generic question can give you a very incomplete picture.
The platform changes the answer too
There is no single AI search result. ChatGPT, Claude, Gemini and Google AI Overviews use different systems and sources, so a company can show up prominently in one and be absent from another.
Google AI Overview availability changed with the question
Share of Google searches in the study where an AI Overview appeared.
Across all 486 Google opportunities, no AI Overview appeared 37.4% of the time. For Google, the first visibility question is sometimes not which company makes the answer. It is whether an AI answer appears at all.
A simpler way to think about it
AI search builds a new shortlist for each question
A better way to think about AI visibility is as a changing shortlist. AI systems assemble a set of businesses that fit the question in front of them. Change the question, location or platform and the shortlist can change too.
Think about recommending a restaurant to a friend
If a friend asks where they should go to dinner, you probably have a loose shortlist of places you would recommend. It is not a permanent ranking. The list changes as you try new places, have a great meal somewhere, have a bad experience somewhere else, or learn something new.
AI systems are not having meals or personal experiences like a person. Their answers change for different reasons, including the question, platform, location, sources they pull from and changes to the AI system over time. But the analogy captures the important part: the shortlist is rebuilt around the question in front of it.
The useful goal is repeated consideration. You want the business to make sense for enough relevant questions that it keeps showing up as buyers get more specific.
What businesses can influence
What makes a business more likely to show up in AI search: The Four Cs
We keep coming back to four areas: Consistency, Clarity, Credibility and Coverage. They are all things a business can actually improve. The clearer and more credible that picture is, the easier it is for both people and AI systems to understand what the company does and when it may be relevant. Our AI search optimization guide explains how this connects to AI SEO, GEO and AEO.
Consistency
Does the same basic story hold together across your website and the rest of the web? Your name, services, locations, leadership, category and core facts should reinforce one another rather than conflict.
Clarity
Can someone quickly understand who you are, what you do, who you serve and where you operate? Make your services, markets, locations and areas of expertise obvious instead of making buyers or AI systems piece the story together.
Credibility
What supports the claims you make about your business? Earned media, reviews, respected industry sources, associations, expert contributions and customer proof all help establish credibility.
Coverage
Where does useful, accurate information about your business show up? Your website matters, but so do search results, news coverage, directories, industry sites, reviews, social platforms, forums and other sources people and AI systems may use.
This goes beyond SEO. Search fundamentals matter, and so do PR, useful content, reputation, third-party coverage, reviews and accurate local information. All of those help establish what a business does and whether it is credible enough to recommend.
What businesses can do now
How businesses can improve their presence in AI search
You cannot control exactly which businesses an AI system recommends. You can make sure the basics are working in your favor: your services and locations are clear, important facts match across the web, and credible third-party sources back up what you say. Then test whether you start appearing more often for relevant questions.
Use realistic buyer questions across broad discovery, direct recommendation and specific needs.
Check ChatGPT, Gemini, Claude and Google separately. They do not return the same businesses.
Make sure important third-party profiles and your own site agree on what you do, where you operate and how you should be categorized.
Publish useful material about the problems you solve, the services you provide, the markets you serve and the work you know best.
Earn credible media coverage, reviews, industry references, expert mentions and partnerships that support what you say about your business.
Correct outdated, conflicting or missing information on the sites and profiles where people are likely to find you.
Re-run a stable set of buyer questions and watch how often your business appears, across how many questions and on how many platforms.
AI visibility is one part of how buyers discover and evaluate you. Connect it to leads, sales conversations and revenue instead of treating visibility as the end goal. We dig into that further in AI Visibility, Attribution and What Comes Next.
How to use the data
Track how often and where you show up
One overall visibility score can hide important differences.
A business might show up often on ChatGPT but rarely on Gemini, or for broad questions but not more specific ones. Track visibility by platform, question type and market, then watch whether your presence gets broader and more consistent over time.
We also saw AI systems recommend businesses that did not fit neatly inside our starting list. At least 10.1% of recommendations pointed to local businesses missing from that list, while another 5.1% pointed to Florida businesses outside the exact target city.
Business categories are messy, service areas overlap, office structures vary and no database is complete. Make your location, service area, category and core services easy to verify wherever people are likely to find you.
Use an overall visibility score as a summary, not the whole story. Break it down by the questions, markets and platforms that matter to your buyers, then watch whether your presence broadens over time.
What we took from the study
AI search is becoming part of the buyer’s research process
Before someone searches for your company by name, they need a reason to find it, understand it and trust it. AI search is becoming one more place where that happens.
The work behind that is familiar: accurate information, useful expertise on your site, independent coverage and reviews, and enough consistency that a buyer or an AI system can tell what your company actually does.
Our study also changed how we think about measurement. Run the same set of realistic buyer questions across several platforms and watch whether your company appears more often and in more situations over time.
If you are tracking AI search, put more weight on repeated appearances across relevant questions and platforms than on any single result.
Florida AI Visibility 2026
See the complete statewide study, city-by-city findings and methodology behind this analysis.
Florida AI Visibility Research
Browse the broader research program, including city and industry studies across Florida.
AI Search Optimization Guide
Use this for the terminology: AI search optimization, AI SEO, GEO, AEO and how the pieces fit together.
Want to see where your business fits?
Run our free AI visibility check for a quick snapshot, or see how Evoke helps businesses become easier to understand, trust and recommend in AI search.
Run the Free AI Visibility CheckExplore AI Search Optimization Services
Methodology questions
How many businesses were included in the study?
The study included 28,488 businesses across nine exact Florida cities and 18 industries.
Which AI platforms were tested?
ChatGPT, Claude, Gemini and Google AI Overviews. Each platform was tested independently using the same three buyer-style question types.
What counted as AI visibility?
We counted a business as visible when we could clearly match it to a recommendation in at least one tested search. Businesses outside our starting list or outside the exact target city were tracked separately.
Will the same businesses always appear?
No. AI systems, the sources they use, the wording of the question and the products themselves change. Repeated testing gives you a much better picture than one result.
