How the AI visibility score is worked out.

Visible Queries asks AI platforms the questions your buyers ask, grades every answer the same way, and turns the pattern into one score out of 100. Here is exactly how, in plain words.

Last updated November 10, 2026Every point traces back to a real answer
  1. 1Read the siteWhat you sell, who buys it, where you operate and who you compete with. You confirm it before anything runs.
  2. 2Write buyer questionsWorded for your category, your buyer types, your competitors and your cities.
  3. 3Ask AIClaude, ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode, with web search on, the way a buyer would ask.
  4. 4Grade every answerWho was named, in what order, how you were described, and any wrong facts.
  5. 5Score the patternA weighted score out of 100, a score per platform, and the numbers behind it.

What we askA mix of questions that leave you out, so we can see if AI brings you up on its own, and questions that name you, so we can see what it says about you.

Leaves you out

Best-of lists

“What are the best CRMs for real estate teams?”

Leaves you out

Buyer questions at every stage

“Why do my agents keep losing track of leads?”

Leaves you out

Alternatives to competitors

“What are good alternatives to Keyhaven?”

Leaves you out

Local questions, by city

“Best real estate CRM for brokerages in Austin?”

Names you

Head to head

“Brightlane or Keyhaven for a 15-agent team?”

Names you

Facts about you

“How much does Brightlane cost?”

Names you

Reputation

“Is Brightlane reliable? What do users complain about?”

Examples use Brightlane, a made-up real estate CRM. Local questions are added for businesses that serve specific cities. Questions for every kind of business follow the same pattern, worded for that industry.

Where we askThe platforms buyers use to research before they buy.

Every question goes to Google AI Overviews and Google AI Mode. The questions that matter most, the ones closest to a purchase, also go to Claude, ChatGPT, Gemini and Perplexity. For these AI assistants we turn web search on and set a location in the United States or Canada, so answers reflect what a buyer in your market would see today, not only what a model memorized during training.

For Google AI Overviews and Google AI Mode we run real Google searches. Google doesn't give an AI answer for every search, so when there is none, that question is left out of that platform's score instead of being counted as zero.

How each answer is gradedEach answer is read by an AI grader with a strict rubric. Whether your site is linked is checked in code against the answer's sources.

Example answer“Best CRM for a small real estate brokerage?”

“Popular choices include Keyhaven for lead routing and ListPilot for transaction tracking. Brightlane is easy to set up and syncs with most MLS feeds.”

  • Named? Yes, third
  • Linked? No
  • Framing Positive
  • Competitors Keyhaven, ListPilot
Grade: Named, not first, not linked

What makes up the score

How much each part counts, out of 100

Real estate80%
Share of voice10%
Head to head10%

Real estate points per answer

For questions that don't mention you, averaged by question weight

Named first and linked to your site100
Named and linked85
Named first80
Named60
Linked, not named50
Not there0

Head to head

100 if AI recommends you over the competitor, 50 if it stays neutral, 0 if it picks the competitor.

Facts about you

Checked for wrong facts and shown in your report as accuracy. Not part of the score: knowing you when asked by name doesn't mean AI recommends you.

Reputation

How favorably AI describes you when buyers ask if you're any good. Feeds the bottom line in your report. Not part of the score.

How the score is worked outNot every question counts the same. Questions closer to a purchase, and questions that describe a specific buyer, count for more.

Buying stage

How close the buyer is to a decision

Not looking yet×0.5
Noticing a problem×0.75
Learning×1
Comparing options×1.5
Choosing×1.5

How specific the question is

Specific questions are closer to how real buyers ask

General×1
For a type of buyer×1.25
Buyer and situation×1.5

Three parts, one score

Each rate is averaged by question weight, then the parts are combined

  • Real estateHow much room you get in each AI answer to a question that doesn't mention you: named, named first, linked to your site. Every answer earns points, and the best answers earn the most.
  • Share of voiceYour mentions out of all brand mentions. Full points at 30% or more, since in a market of about five brands that share means you lead.
  • Head to headWhen a buyer compares you with a competitor, does AI recommend you?

So a business named in a third of answers and linked in a sixth scores in the 20s, not the 40s. If a part has no answers, it is left out and the other parts are scaled up. Each platform and each city gets its own score, worked out the same way from its own answers.

What the score means

Market Leader 81–100
AI names you first in most answers and links to your site. You are the default recommendation for the questions your market asks.
Prominent 61–80
AI names you in most of the questions your buyers ask, often first. A few gaps remain where you are missing or a competitor is picked ahead of you.
Partly Visible 41–60
You show up in some answers, but you are missing from many of the core questions your market asks, or competitors are named ahead of you.
Overlooked 21–40
You show up sometimes, but you are missing from most of the questions your market asks, or named behind competitors, so buyers likely drop you from consideration.
Invisible 0–20
AI rarely or never names you, so buyers researching with AI don't know you exist.

The other numbers in your reportWorked out from the same graded answers.

Mention rate
The share of answers that name you, out of the questions that don't mention you by name.
Citation rate
The share of those answers that link to your website as a source.
First mention
The share of those answers where you are the first brand named.
Share of voice
Your mentions divided by all brand mentions, yours plus every competitor's, across those answers. In a market of about five brands, 30% or more means you lead.
Accuracy
The share of answers about your company that contain no wrong facts, checked against what your own site says.
Search demand
Estimated monthly Google searches for each topic, as a stand-in for how often buyers ask. AI platforms don't publish question volumes, so this is always an estimate, and it never changes your score.

What to keep in mindNo measurement of AI answers is exact. These are the limits of ours.

Answers change from run to run
The same question can get a different answer an hour later. That's why we ask many questions and score the pattern, not any single answer. Treat your score as a baseline and re-check monthly.
It's a snapshot
AI platforms update their models and search indexes often. A score reflects the answers on the day of the check.
Questions are our best estimate of what buyers ask
We write them from your website and the buyers and competitors you confirm. Real buyers phrase things in endless ways, so the questions are a representative sample, not every possible prompt.
The grader can make mistakes
An AI grader reads every answer with a strict rubric, and links are checked in code. It is accurate on clear-cut answers and can misread a few edge cases.
US and Canada, in English
Questions are asked in English from locations in the United States and Canada.
Citing this page

Visible Queries. “How the AI visibility score is worked out.” Updated November 10, 2026. https://www.visiblequeries.com/methodology

Free · a few minutes · just a URL

See your own score.

Get one free score, side by side with your competitors. Just enter your URL.

Real buyer questionsSee who AI recommends insteadNo setup, no credit card