Free AI Visibility Snapshot
When customers ask AI who to hire, are you on the list?
We measure how your business appears across a selected set of AI-powered search and recommendation tools — today ChatGPT, Gemini, Claude and Perplexity. We ask the questions your customers ask, twice, and send you a one-page answer: which businesses appear, whether you do, and how the results vary. Free, in five business days.
Real data from our first benchmark: one company's presence across 10 buyer questions × 4 AI tools (ChatGPT, Gemini, Claude, Perplexity) × 2 runs.
The problem
AI-powered search and recommendation tools are forming your customers' shortlist. Most businesses have no idea whether they're on it.
More buying decisions now begin with a question to an AI tool — a chat assistant, or the AI answer at the top of a search page — instead of a list of links. The tool doesn't return ten blue links; it returns a handful of names and tells the customer why.
That handful is a consideration set. Some businesses appear in it over and over. Others, including well-established ones, barely appear at all. The difference is measurable, and it's invisible from inside your own company.
- of eligible answers recommended the most visible company in our benchmark
- 71.6%
- recommended an established competitor with comparable regional presence
- 13.5%
- of the time a tool named the same company first when asked the identical question again
- 53%
Why one search isn't an audit
Typing your name into ChatGPT once tells you almost nothing.
AI recommendations are not a ranking. They're a distribution. The same question, asked the same way, produces different shortlists depending on which tool you ask, how the question is phrased, and simply when you ask it. A single screenshot is an anecdote. A benchmark needs repetition — our first one collected every question twice, as two passes a day apart, which is enough to show variation but not enough to call any rate stable.
Run-to-run overlap, by assistant
How much the recommended names overlapped when we asked each tool the identical question twice: names present in both runs, divided by all names across the pair, averaged over that tool's eligible pairs (34 pairs in total).
- ChatGPT65%
- Perplexity64%
- Claude31%
- Gemini25%
Differs by tool
In our benchmark the top company appeared 18 times on Claude and 4 times on Gemini. Another company appeared 16 times on Perplexity and zero on ChatGPT.
Differs by question
"Who installs heat pumps?" and "Who should I call if my AC dies?" surfaced noticeably different sets, even on the same tool.
Differs between runs
Gemini's answer sets overlapped only ~25% run-to-run. ChatGPT's overlapped ~65%. Any single answer is partly noise.
How we measure
A controlled measurement, not a vibe check.
- 01
Freeze the questions
We write the questions the way real customers ask them for your category and market, and freeze them before collecting anything. No company names, no leading language, and no cherry-picking questions a business happens to do badly on. These are designed buyer questions, not a measured sample of what people actually search.
- 02
Ask every tool in the set, repeatedly
Each question goes to the selected tools — currently ChatGPT, Gemini, Claude and Perplexity — in fresh sessions, with personalization turned off where the platform allows. We record the visible model or mode and anything we can't verify. The whole set is collected at least twice — two passes, not two independent samples. First answer only: no regenerating, rewording or follow-ups.
- 03
Preserve everything
Raw responses, citations, map cards and technical failures are all kept. A failed request is logged, not quietly re-run; a replacement attempt never counts as an extra observation. Every number you receive traces back to a specific captured answer.
- 04
Extract and benchmark
We code every business a tool positively presented as an option, once per answer. Recommendation presence = answers recommending the company ÷ all eligible answers. Several companies can appear in one answer, so percentages don't sum to 100 — this is not market share. We report it per tool, with each tool's own denominator, plus first-position counts and run-to-run overlap.
- 05
Diagnose the gap
This is a separate step from measuring. We compare you against the most visible competitors on observable signals: the sources the tools cited, local entity consistency, independent corroboration, and owned-site coverage. An uncaptured source is unknown, not absent; a source appearing beside a company doesn't prove it supported that company. We report correlations as correlations and look for counterexamples on purpose.
- 06
Prioritize and re-measure
Our evidence ladder is observation → correlation → hypothesis → proposed intervention → tested effect, and each step needs more evidence than the last. You get 3–5 testable interventions, each with its hypothesis and how we'll know if it worked. The same frozen questions run again afterward; a before/after change is a starting point, not proof, so we design the comparison before anything is changed.
Benchmark
What this looks like with real data.
Our first benchmark measured residential HVAC companies in a Northern Virginia market, collected October 4–5, 2026: 10 frozen customer questions × 4 AI tools (ChatGPT, Gemini, Claude and Perplexity) × 2 passes, 80 observations, 74 eligible after excluding six technical failures. Company names are withheld here; the full methodology and raw responses exist for every figure.
Platform × intent presence
Select a company to see where it appeared.Recommendation share, by tool
| Company | ChatGPT | Gemini | Claude | Perplexity | Share |
|---|---|---|---|---|---|
| Company A | 17 | 4 | 18 | 14 | 71.6% |
| Company B | 14 | 3 | 13 | 7 | 50.0% |
| Company C | 0 | 9 | 1 | 16 | 35.1% |
| Company D | 1 | 3 | 0 | 6 | 13.5% |
Counts are positive written recommendations across eligible answers. Share = mentions ÷ 74 eligible observations. Several companies can appear in one answer, so shares don't sum to 100% and are not market share.
Company A and Company D are both established regional providers. One appeared in roughly seven of every ten eligible answers; the other in about one in seven — and almost never on two of the four tools. We don't yet know why. That is exactly the question the diagnostic phase is built to answer, without guessing.
What this benchmark can and can't say
- One industry, one locality, ten designed questions, four consumer products (ChatGPT, Gemini, Claude, Perplexity), two passes. Google Search's AI features were not part of this benchmark. Results can change with time, geography, accounts, product versions and retrieval.
- Two passes show that answers vary; they don't estimate a stable rate, and we make no statistical-significance claims.
- Sessions were fresh with personalization off where offered, but not every condition could be independently verified; we record what we saw and what we couldn't.
- It establishes a measurable gap. It does not establish why, and no visibility improvement has been demonstrated by anyone — including us.
Want to see this for your business?
The free Snapshot is a smaller version of exactly this — your category, your market, your competitors.
The offer
Start free. Go deeper only if the numbers say you should.
Three stages, each one a complete deliverable on its own. Most businesses should start with the Snapshot; it tells you whether there's a problem worth paying to understand.
- Free01
Snapshot
Find out if you're on the list.
- 3 real customer questions for your category and market
- Asked on ChatGPT, Gemini, Claude and Perplexity
- Every question run twice — 24 planned first-answer observations, not one screenshot. We report how many were usable and any failed attempts.
- One-page result: who's recommended, whether you are, on which tools
- Delivered within 5 business days. No call required.
- Scoped per engagement02
Benchmark
The full picture, measured properly.
- 10+ frozen questions across every buyer intent
- The same selected AI tools × repeated runs
- Competitive recommendation map — every competitor that appeared, and where
- Platform × intent matrix for your business
- Source & citation analysis: what the tools actually relied on
- Run-to-run stability and first-position frequency
- Scoped per engagement03
Diagnostic + Re-measurement
Why the gap exists, and whether it moved.
- Side-by-side comparison with the most visible competitors on observable signals
- Visibility gap diagnosis — stated as correlations, not causes
- 3–5 prioritized, testable interventions, each with a hypothesis and measurement plan
- The same frozen Benchmark re-run after changes, so you see what actually moved
Our guarantee is about the work, not the algorithm.
If your Snapshot shows you nothing you didn't already know, we'll tell you that plainly — and you've spent nothing. On paid work, you get the deliverable described or you don't pay. What we will never guarantee is a ranking, a placement, or a percentage gain, because no honest measurement firm can.
Why is the Snapshot free?
Because every Snapshot is also research. We're building benchmarks across categories and markets, and your three questions add to a dataset nobody else has. You get an answer; we get a data point. If the answer shows a gap worth closing, we'll tell you what a full Benchmark would measure and what it costs.
Why this isn't SEO
Rank tracking assumes a ranking exists. Here, it doesn't.
We don't claim to know how any AI tool ranks businesses, and we're skeptical of anyone who says they do. What we can do is measure rigorously, compare systematically and test honestly.
Who this is for
Businesses where being on the shortlist is the whole game.
Our method is category-agnostic. It works anywhere a customer might describe a problem to an AI tool and act on the names it returns.
Local services
HVAC, plumbing, roofing, electrical, pest control, landscaping, remodeling
Professional practices
Dentists, orthodontists, med spas, law firms, accountants, financial advisors
Hospitality & retail
Restaurants, hotels, specialty retail
B2B & software
SaaS products, agencies, consultancies, professional services
This is not for you if…
- You want someone to guarantee you'll show up in ChatGPT or in Google's AI answers. Nobody can, and we won't pretend.
- You're looking for an SEO agency with a new label. We measure AI recommendations directly; we don't sell backlinks.
- Your customers never describe a problem to an AI tool and act on the answer. If nobody asks AI about your category, there's nothing to measure yet.
- You need something by Friday. Controlled measurement takes days, not hours, because the repetition is the point.
FAQ
Reasonable questions.
What's the catch with the free Snapshot?
There isn't one. It's three questions instead of ten, and one page instead of a full report. It's also research for us — we're building benchmarks across categories. If the Snapshot shows a gap, we'll tell you what a full Benchmark would cost. If it doesn't, we'll say so.
Can you guarantee my business will show up in ChatGPT?
No — not in ChatGPT, and not in Google's AI Overviews either — and we'd be suspicious of anyone who says they can. These tools don't publish how they choose businesses, and the answers vary between runs. What we guarantee is a rigorous measurement of where you stand, a systematic comparison against the businesses that do appear, and an honest re-measurement after any changes.
Isn't this just SEO with a new name?
The inputs overlap — reputation, citations, a clear web presence — but the output is different. There's no ranked page to climb. There's a short list that forms differently on each tool and each question. We measure that list directly rather than inferring it from search rankings.
Why ask the same question twice?
Because the answer changes. In our first benchmark, Gemini's recommendation sets overlapped only about 25% between identical runs. If we only ran once, we'd be reporting noise as fact. Repetition is how we separate a real pattern from a lucky screenshot.
What does "recommended in 71.6% of answers" actually mean?
That the company was positively named as an option in 53 of the 74 usable answers in our first benchmark. Several companies are usually named in one answer, so these percentages don't add up to 100 and aren't market share. First position means it was named first — not that the assistant declared it the best.
Which AI tools do you measure?
A selected set, not every AI-powered surface. Today that is ChatGPT, Gemini, Claude and Perplexity in their consumer-facing modes — every figure on this page comes from those four. We are working out how to measure Google Search's AI features (AI Overviews and AI Mode) under the same controlled protocol; they are not yet part of the Snapshot or the benchmark above, and this page will say so when they are.
Do you also do the fixes?
The Diagnostic ends with a prioritized plan. Implementation can be done by you, your existing vendors or us — but every recommendation is tied to a hypothesis and a measurement, so whoever does the work, you'll know whether it moved anything.
What does the paid work cost?
Scope depends on category, market and competitor count. Start with the free Snapshot; if there's a gap worth understanding, we'll come back with a specific plan and price.
Free Snapshot
Find out if you're on the list.
One minute to request. Five business days to your answer.