No black box. Here is exactly what it measures.
A visibility score you cannot check is a number somebody made up. Everything below describes the method — the questions, the models, the weights and the limits — so you can decide for yourself whether the figure the console gives you means anything.
Run the scan01
What we ask
Fifteen questions a buyer would actually type, across five angles. Your brand is never one of the words in them.
Three questions each. One phrasing measures a phrasing; five angles measure a market. And because your name never appears in the question, a model cannot agree with a premise we handed it.
- Open recommendation
- The bare question a buyer starts with, with no constraints attached.
- Shortlist & procurement
- The question asked when a list is being drawn up and budget exists.
- Trust & reputation
- Who the model considers safe to vouch for, which is not the same as who it knows.
- Segment fit
- Startup, enterprise, international — the same market, different answers.
- Alternatives & switching
- Where a model sends someone already unhappy with a competitor.
Shown for fintech & payments. Whichever category you pick is swapped into the highlighted words — the rest is sent exactly as written.
- 01What are the best fintech & payments companies right now?
- 02Which fintech & payments companies would you recommend?
- 03Who are the leading providers in fintech & payments?
- 04I need to shortlist fintech & payments partners. Who should I be looking at?
- 05Which fintech & payments vendors would you recommend to a mid-market B2B buyer?
- 06We are running a procurement process for fintech & payments. Which suppliers should be on the list?
- 07Who are the most trusted providers in fintech & payments?
- 08Which fintech & payments companies have the strongest reputation with enterprise clients?
- 09Which fintech & payments providers have the best reviews from their customers?
- 10Which fintech & payments companies are best for a startup on a limited budget?
- 11Which fintech & payments providers work best with large enterprises?
- 12Which fintech & payments companies are best for a business expanding internationally?
- 13What are the main alternatives in fintech & payments worth comparing?
- 14Who competes with the biggest names in fintech & payments?
- 15If I am unhappy with my current fintech & payments provider, who should I move to?
02
Who we ask
Three models answering from memory, two answering from live search. They measure two different things.
ChatGPT · Claude · Gemini
Asked with web search off, so the answer comes from what the model absorbed during training. This is the question “does the industry talk about you enough that a model learned your name?” It moves slowly and it is the harder one to fake.
45 answers · this is the free half
Perplexity · Google AI Overview
Asked with retrieval on, so they read the web while answering. This is the question “are your pages findable and quotable today?” It moves fast, and it is the one you can influence this quarter.
Unlocked with an email · costs us money to run
Most tools report one number and call it AI visibility. But a brand can be famous and unreadable, or unknown and perfectly structured, and those two problems have opposite fixes. Being absent from memory but present in retrieval means your content is fine and nobody is citing it. Being present in memory but absent from retrieval usually means a robots.txt line is undoing years of reputation.
03
What we count
Three things per answer, and one of them is the rank you held.
Named or not
Whether the answer contains your brand at all. Aggregated, this is your share of voice — the headline percentage.
Where you ranked
A model that lists you eighth is not recommending you. Average rank is reported next to the share of voice for exactly that reason.
Who was named instead
Every other company in every answer, tallied. That list is the competitive picture, and it is usually the part that changes the conversation internally.
04
What we score
A hundred points across your own site, weighted by how much each one actually moves.
Technical readiness
Checked live against the domain you enter. Every point in your score traces to one of these rows — there is no hidden component.
- AI crawlers allowed in robots.txtGPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot. Blocked here and nothing else on this list matters.25 pts
- Cited on third-party authority sitesWhether anyone but you says you exist. The heaviest signal, and the slowest to move.25 pts
- Organization schema markupLets a model resolve who you are rather than infer it from prose.15 pts
- About page with named peopleAnonymous companies are hard for a model to vouch for.10 pts
- Content published in the last 6 monthsRetrieval pulls what is current, however good the older pages are.10 pts
- Structured FAQ schemaA marked-up answer can be lifted whole; a paragraph gets paraphrased away.10 pts
- HTTPSTable stakes, scored because its absence is disqualifying.5 pts
- llms.txtListed and deliberately not scored — no major AI system is known to read it, so its absence costs nothing.not scored
05
What you get
The finding is free. The diagnosis costs an email — and nothing else.
- Your share of voice across the three models
- Your average rank in the answers that named you
- Every question we sent and every result, in a table
- The competitors recommended in your place, ranked
- Perplexity and Google AI Overview presence
- The full technical scorecard against your domain
- A prioritised plan, ordered by what the scan found
One email, used to send you the report and to reply. We do not sell or share it, and you can ask us to delete it.
06
What it does not do
Stated here rather than discovered later. Every one of these is a real limit of the method.
- Models are asked once each, on the day you run it. Answers drift week to week, so a single scan is a snapshot rather than a trend.
- We ask in English, from our servers. A model that personalises by location or account history may answer your buyers differently.
- A brand sharing its name with something more famous will pick up mentions that are not about you. The audit trail is there so you can see when that has happened.
- Google AI Overview is read through a search-results provider, not an API Google publishes. It is reported as not measured rather than guessed at when that lookup is unavailable.
Measuring it is the easy half.
The scan tells you where you stand. Moving it takes the content, the citations and the structure that put you in the answer — which is the work we do.