The State of AI Visibility, 2026
Original data from 127 live AI category audits: how many businesses AI actually names when someone asks it to recommend a provider — and what happens to everyone it doesn't.
Short answer: Across 127 live category queries run through Google's Gemini 2.5 Flash — spanning 9 service-business verticals and 15 US metros — the AI named between 6 and 8 businesses per query, averaging 7.97, and never once declined to name anyone. There is no "no clear leader" answer in the AI layer. There is a shortlist of about 8 names, and there is everyone else. Get your free AI-visibility audit to see which side you're on.
Three headline findings
What we actually measured
We ran 127 real prompts through an AI answer engine, one per (industry, city) pair, asking it to name businesses the way a buyer actually would — "who are the best [vertical] in [city]." We logged exactly which names came back and how many. This is not a survey or a model of behavior. It is the raw output of the engine, captured and counted.
The 127 audits cover 9 verticals — law firms, personal injury lawyers, dental implant clinics, cosmetic dentists, chiropractors, med spas, HVAC companies, plastic surgeons, and real estate agents — across 15 US metro markets including Miami, Orlando, Tampa, Naples, Fort Lauderdale, Jacksonville, Atlanta, Charlotte, Dallas, Austin, Denver, Scottsdale, and San Diego. Most verticals were checked in 14-15 cities each; real estate agents were checked in 8.
Finding 1: AI runs an 8-seat shortlist, not a ranked list
The single clearest pattern in the data: the engine consistently names 6 to 8 businesses per category query, never more, never fewer than 6, and averaging 7.97. It behaves less like a search results page with ten blue links and more like a fixed-size shortlist — a finite set of seats that get filled every time, regardless of how many real businesses actually compete in that market.
That has a blunt implication. If your market has 40 competing chiropractors and the AI only ever names 8, being "pretty well known locally" is not enough. You are either occupying one of roughly 8 seats in every relevant query, or you are functionally invisible in that channel — no matter how strong your Google Business Profile or your organic rankings look.
Finding 2: the AI never says "I don't know"
Zero of the 127 queries came back empty or hedged. Every single one produced a confident, fully-populated list of named businesses. Classic search has a graceful failure mode — a thin results page, a "no reviews yet" listing, an honest gap. Generative answer engines do not show that gap. They commit to a shortlist every time, even in categories where the honest answer might be "it depends."
This matters because it removes the excuse of "well, AI probably doesn't have a strong opinion on my category yet." It does. It has an opinion on every category we tested, all the time, and that opinion either includes your brand or it doesn't.
Finding 3: national names crowd out local ones
Looking at which specific businesses repeated across multiple markets, a pattern of recognizable, high-authority national and multi-market brands dominating the shortlist emerges. Morgan & Morgan appeared in 8 of the law-related market checks — more than half. Keller Williams Realty appeared in 4 of 8 real-estate markets, Coldwell Banker Realty in 3. Akerman LLP and Gunster, two large multi-market Florida law firms, each repeated across several cities.
The businesses that show up once, tied to a single city, are almost always independents and regional specialists — the exact category of business that has historically competed well on local SEO and word of mouth, but that this data suggests is losing shortlist seats to household names with broader digital footprints and heavier corroboration across the web.
| Business | Vertical | Markets it appeared in (of dataset) |
|---|---|---|
| Morgan & Morgan | Law firms / personal injury | 8 |
| Precision Chiropractic | Chiropractors | 6 |
| Compass | Real estate agents | 6 |
| Akerman LLP | Law firms | 5 |
| Keller Williams Realty | Real estate agents | 4 |
| Radiance Medspa | Med spas | 4 |
| Coldwell Banker Realty | Real estate agents | 3 |
| Gunster | Law firms | 3 |
Note: "Precision Chiropractic" and "Radiance Medspa" are generic-sounding names that recurred verbatim across unrelated cities in the raw model output — a sign the engine may sometimes generate plausible-sounding but non-verified names for thinner categories. We are flagging this as a data-quality caveat, not smoothing it out.
Methodology
Collection method: a scripted audit tool queried Google's Gemini 2.5 Flash model once per (vertical, city) pair with a natural-language prompt asking the model to name businesses in that category and location, the way a consumer would phrase the question. Every response was logged verbatim as structured JSON — the vertical, the city, the model used, the full list of named businesses, and a count.
Sample: 127 completed audits across 9 verticals and 15 US metro markets, collected as part of Acromatico's AI Visibility Finder tooling.
Engine coverage: this dataset currently reflects one engine — Google Gemini 2.5 Flash. It does not yet include ChatGPT, Perplexity, or Claude.
What is real vs. placeholder: every statistic above (127 audits, 9 verticals, 15 cities, 7.97 average names, the 6-8 range, the 100% full-response rate, and every named-brand repeat count) is computed directly from the logged audit files — nothing here is estimated or fabricated. We separately maintain a multi-engine citation-tracking table (Claude and Perplexity, in our internal SEO rank-tracker) built for this exact purpose; as of publish it has not yet accumulated data, so multi-engine citation-rate comparisons are intentionally left out of this report rather than filled in with a guess. We will publish that breakdown once it has a real sample.
What this means if you run a business
If your category has an AI shortlist of roughly 8 names and you are not on it, you are not "ranking lower" — you are absent from the only page the buyer sees. There is no AI equivalent of page two. The audit above is a live demonstration of exactly this mechanic in service categories most likely to overlap with our own client base; the same 8-seat pattern is what we check for, vertical by vertical, market by market, when we run a free audit on your brand.
Want to see it for your own category and city? Run the same check yourself with our free AI Visibility Finder — it takes 20 seconds and shows you the real, live answer. If you're missing from the list, our AI Visibility Sprint is the fastest way to fix the underlying signals.
Frequently asked
How many businesses did you audit for this report?
We ran 127 live category audits across 9 service-business verticals and 15 US metro markets, each one a real query asking an AI engine to name providers in that category and city.
Which AI engine was used for this dataset?
This dataset was collected with Google Gemini 2.5 Flash. We are expanding collection to ChatGPT, Perplexity, and Claude and will publish those numbers once the sample size is large enough to report responsibly.
How many businesses does AI typically name in a category search?
Across all 127 audits, the answer engine named between 6 and 8 businesses per query, averaging 7.97. It never named zero and it never named more than 8.
Does AI always give a clear answer, or does it sometimes say it doesn't know?
In this dataset, 100% of the 127 queries returned a full named shortlist. The engine never declined to answer or said it lacked information. It always committed to a list, which means every category search has some businesses inside the shortlist and everyone else is left out.
Is this dataset a one-time snapshot or an ongoing tracker?
It started as a one-time audit pull. We are building it into a continuously updated tracker inside our SEO rank-tracking system, and this page will be updated as new data comes in.
Find out if you're one of the 8
We'll run your category and market the same way, live, and show you in plain language whether AI names you — and exactly what's in the way if it doesn't.
Data collected via Acromatico's AI Visibility Finder tooling · Last updated: July 8, 2026