We Ran 700 AI Visibility Checks in One Texas Market. Here's Who AI Actually Recommends.

Most claims about AI visibility are vibes.

Someone asks ChatGPT a question once, screenshots the answer, and writes a strategy post about it.

One check. One day. One phrasing.

That is not measurement.

So we measured.

Short answer

Between July 7 and August 3, 2026, 6Signal's tracking system ran 700 automated checks of 12 real buyer prompts across five AI engines and Google Maps, for one plumbing company in Red Oak, Texas — a Dallas–Fort Worth market.

The results, by engine:

Same company, same prompts, same market. The zero is the finding.

On Google Maps, a separate 25-point grid scan across the market for "best plumber red oak tx" found the company present at all 25 grid points, at an average position of roughly 5.5.

Same company. Same market. Same buyer questions. Five completely different outcomes.

That spread is the story.

The method — so you can judge the data

Disclosure first: the company is X-Act Plumbing, a 6Signal client. The numbers are published with the client's permission, and they come from our own tracking system — the same one that powers the AI Visibility Audit.

How the tracking works:

12 prompts, written the way buyers actually talk. Not keywords. Questions like "Who is the best plumber in Red Oak, TX?", "I have a burst pipe in Red Oak, TX — which plumber can come out fast?", "Who should I call to replace a water heater in Red Oak, TX?", and "How much does a plumber cost in Red Oak, TX?" — plus brand-verification prompts like "What do reviews say about X-Act Plumbing?"
Six surfaces, checked daily. ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews, and Google Maps.
One binary question per check: is the business named in the answer, or not?
700 checks over four weeks, July 7 through August 3, 2026.

Repetition is the point. Any single AI answer varies by day, phrasing, and user. Patterns across hundreds of checks do not.

If you want to run this kind of testing yourself, the manual version is documented in How to Test Whether AI Recommends Your Business.

What the spread means

Why Gemini says the name 92% of the time

Gemini sits closest to Google's local data — the Business Profile, the reviews, the Maps entity. For a company with a clean local entity record, Gemini has the most confident grounding available, and it shows: 114 of 124 checks named the business.

The lesson generalizes: engines answer local questions from local entity data. The company's local entity signals — consistent name, phone, service area, categories, review volume — are doing the recommending.

Why ChatGPT and Perplexity land in the 70s

Both name the company in roughly seven of ten checks — strong, but less consistent than Gemini. Both engines assemble local answers from what the open web says: directories, review platforms, the company's own site, third-party mentions. That composite is strong for this company but not airtight, and the variance shows up as the missing 30%.

Perplexity's habit of citing sources makes it the most instructive engine to study: when it skips a business, you can usually see which sources it trusted instead. Often that's a directory — the pattern covered in Why Directories and Aggregators Keep Winning AI Answers.

Why Claude names it only 27% of the time

Claude is the most conservative engine in the set about recommending specific local businesses — it more often answers local questions with criteria ("look for licensed, insured plumbers with recent reviews") than with names. that caution shows in the data: 33 of 124.

The takeaway is not "ignore Claude." It's that mention rates are engine-specific, and a strategy tuned only to one engine misreads the field.

The zero: Google AI Overviews

Eighty-four checks. Zero mentions.

Not zero because the company is weak — the same company scores 92% on Gemini and stands on all 25 Maps grid points. Zero because AI Overviews is a different game: it cites indexed web pages that directly answer the query, and it triggers selectively on local-service searches. Winning it is not an entity play; it is a content play — answer-ready pages that say, in plain indexed text, what the Overview needs to quote.

This is the most important finding in the dataset: a contractor can dominate four AI surfaces and be invisible on a fifth. Visibility is not one number. It is six different systems keeping six different scores.

Questions this data raises

Does a 92% mention rate mean 92% of buyers see the name?

No. It means that when our tracked prompts were asked during the window, the engine named the business in 92% of checks. Real buyers phrase things differently, ask from different places, and get personalized results. Treat mention rate as a signal-strength reading, not an impression count.

Is one company in one market enough to generalize?

It is enough to demonstrate the shape of the problem — the same business scoring 92%, 75%, 71%, 27%, and 0% across five surfaces — and that shape is structural, not company-specific. The engines ground their answers differently, so uniform performance across them would be the surprise. Your rates will differ; the spread will not.

Why publish the zero?

Because the zero is where the work is. We don't publish outcomes we can't verify, and we don't hide the surfaces we haven't won. A measurement system that only reports victories is a marketing system.

What moved the numbers that are high?

The boring fundamentals, in the boring order: local entity cleanup first (Business Profile, reviews, citation consistency), machine-readable structure second, answer-ready content third. The full framework is in The 6Signal Visibility Audit: What We Measure and Why, and the measurement discipline in How to Measure AI Visibility Without Lying to Yourself.

What this means for your company

If you're a contractor, three usable conclusions:

You have six scores, not one. Ranking well on Google tells you nothing about ChatGPT. Winning ChatGPT tells you nothing about AI Overviews. Each surface has to be checked on its own.
Mention rates are measurable. "We do AI SEO" is a claim. "Named in 75% of tracked ChatGPT checks over four weeks" is a measurement. Demand the second kind of sentence from anyone you pay — including us.
The recommendation layer is already running in your market. Someone is being named for your trade in your city right now, at some rate, on every one of these engines. The only question is whether it's you — the argument made in full in The Recommendation Layer.

Final answer

Across 700 checks in one Texas market over four weeks, the same plumbing company was named in 92% of Gemini checks, 75% of ChatGPT checks, 71% of Perplexity checks, 27% of Claude checks, and 0% of Google AI Overview checks — while standing on all 25 points of a Maps grid scan.

AI visibility is not one thing. It is six systems keeping six scores, and every one of them is measurable.

Want your six scores?

Get the AI Visibility Audit. The $27 audit runs your company through all six signals — instantly, for your trade and your market — and shows you where you're named, where you're skipped, and what to fix first.

Sources and method notes

All mention-rate figures: 6Signal internal tracking system, 700 automated checks, July 7 – August 3, 2026, published with client permission. Counts by engine: Gemini 114/124, ChatGPT 72/96, Perplexity 99/140, Claude 33/124, Google AI Overviews 0/84.
Maps data: 25-point geographic grid scan, query "best plumber red oak tx," July 2026.
Google Search Central: AI features and your website
Google Business Profile: Tips to improve your local ranking