Skip to content

[Method]

How every answer is made.

Every number in Spotlit comes from the same seven steps. Here they are, with how each metric is counted, what we don't measure, and the limits.

Scan freeSee how it works
Scan log
Sample · fictional brand
  1. 01Read siteDone
  2. 02Buying questionsDone
  3. 03Live web searchDone
  4. 04Answers writtenDone
  5. 05Brands and pagesDone
  6. 06Repeat runsDone
  7. 07Visibility score48
Typical time: about 2 minutes

[01]The pipeline/ seven steps, every scan

From your homepage to a score.

The same steps run on every scan and every weekly re-check.

  1. 01

    Read your siteYour homepage

    We read your homepage to find your brand, what you sell and who buys it. You check it and fix anything we got wrong.

  2. 02

    Pick buying questionsup to 10 buying questions

    Questions buyers ask in your category, plus one about your brand, in your country and language. They are saved, so every re-check asks the same ones.

  3. 03

    Search the live webTop 10 results

    Each question runs a web search at scan time. The answer is built from today's pages, not from what a model remembers.

  4. 04

    Write the answerReal ChatGPT answers

    Our own AI model answers using only those search results, as a ranked list, and keeps every page it used.

  5. 05

    Find brands and pagesNames · places · links

    We find every brand named, its place in the list, and every page cited, then check which of those pages mention you.

  6. 06

    Ask againone live ChatGPT answer per question

    Answers change from run to run, so the key questions run again and two are asked in other words. You see how often you're named across runs.

  7. 07

    Score and compare0–100

    Names and places become a visibility score, next to your competitors on the same answers. Weekly re-checks show what changed.

Steps 03–05 · one question

AI answer
Sample · fictional brand

Which email marketing tools are best for email automation?

Searched the web · read 3 pages

For automation depth, the usual shortlist is: 1. ActiveCampaign 2. Klaviyo 3. Mailchimp 4. Acme Email 5. Brevo ActiveCampaign leads on branching logic, while Acme Email covers the basics at a lower price point.

  • 1g2.com
  • 2capterra.com
  • 3reddit.com

Spotlit reading

Acme Email
Named #4
Rivals named
4
Pages cited
3
Sources AI cited
Sample · fictional brand
  • g2.com6×Missing
  • pcmag.com5×Missing
  • hubspot.com4×Missing
  • mailchimp.com4×Missing

The answer is written only from the pages the search found. Each page is saved, and checked for your name.

[02]The numbers/ how each one is counted

What each number means.

Four numbers carry the report. Each is counted the same way for you and for every competitor, from the same answers.

01 · Metric

Visibility score

How strongly AI names you across your buying questions, from 0 to 100.

How it's counted

Each question earns points for your place in the answer. We add them up, divide by the most you could get (10 per question) and take the square root, so early progress shows. The question with your brand name in it is left out.

Points per question by place in the answer
Your place in the answerPoints
Top 310
4th to 10th6
11th or lower3
Named, no clear place2
Not named0

ExampleNamed 4th in one of three questions: 6 of 30 points is 20%. The square root of 0.20 is about 0.45, so the score is 45.

02 · Metric

Named in

How many answers name you, out of the answers checked.

How it's counted

We look for your brand name and its common spellings (for example “Peec AI”, “Peec” and “PeecAI”). Over a date range it is the answers that named you, divided by all answers in that range.

ExampleNamed in 3 of 8 answers: a visibility rate of 38%.

03 · Metric

Cited pages

Every page the answers were built from, grouped by site.

How it's counted

We save each page's link, count how often each site is cited, and check whether the page mentions you (its title and the text the search returned). A cited site that doesn't mention you, and isn't your own, is a gap.

04 · Metric

Competitors ahead

How many brands scored higher than you on the same questions.

How it's counted

Every competitor gets the same visibility score from the same answers. That covers the up to 5 rivals you choose and, in the dashboard, any brand the answers ranked in at least 2 answers, so a rival you didn't add can't hide.

ExampleYou score 48. Your rivals score 63, 61 and 37: two competitors ahead.

Share of voice
Your mentions as a share of all brand mentions in the answers.
Average position
Your mean place when an answer ranks you. Places move often; read it as a tendency.
Gap score
0–100 for a site you're missing from: mostly how often AI cites it, partly how many questions use it.
Repeat rate
How often you're named when the same key question is asked again.

[03]Scope/ what we measure, what we don't

What we measure. What we don't.

A mention, a citation, a visit and a customer are different things. We say which one we saw.

We measure

  • Yes:Whether the answer to each buying question names you, and in what place
  • Yes:Every page each answer cites, and whether that page mentions you
  • Yes:Which competitors are named on the same questions, and how often
  • Yes:How stable that is when the same question is asked again
  • Yes:What the model says about your brand from memory alone, with no search
  • Yes:What changed since your last scan
  • Yes:Whether AI crawlers can reach your site

We don't measure

  • No:What ChatGPT, Perplexity, Claude, Gemini or Google AI Overviews show inside their own apps
  • No:Real people's chats, or how often people ask a question
  • No:Traffic, leads or revenue that came from AI
  • No:Sentiment: whether AI speaks well or badly of you
  • No:Results per country inside the AI apps

[04]Known limits/ read before you rely on a number

The limits, said plainly.

Every tool in this category has them. These are ours, and what each one means for how you read a report.
  1. 01

    Answers change from run to run

    Ask the same question twice and the list can change. Places move more than names do.

    What it means: Read one scan as a snapshot. Trust the trend over several weekly scans more than any single number.

  2. 02

    The engine is simulated

    Answers come from the real ChatGPT app, collected live. Real AI apps use their own models, searches and memory, so they can name different brands.

    What it means: Use Spotlit to see which pages shape answers in your category and what to fix, not as a copy of what ChatGPT shows.

  3. 03

    Country is a search lean, not a local app

    Choosing a country makes the web search favour that country's pages, and the answer comes back in your language.

    What it means: It is not a reading of ChatGPT as seen from that country.

  4. 04

    Small question sets move in big steps

    A scan asks up to 10 buying questions. One answer changing can move the score by several points.

    What it means: Add the questions your buyers really ask, and compare like with like: same questions, same market.

  5. 05

    Names are matched as text

    We look for your brand name and its common spellings. A rare misspelling can be missed, and a brand named after a common word can be over-counted.

    What it means: Open the answer behind any number. Every one is saved in full, with its pages.

  6. 06

    Fan-outs are predicted

    The searches shown behind a prompt on the Fan-outs page are predicted by our model, not captured from an AI app.

    What it means: They are labelled “Predicted” everywhere in the app.

[05]Review/ dated

Last reviewed

When the method changes, the changelog says so.

This page is checked against the scan code. If a step, a formula or a limit changes, we update it here and add a dated entry to the changelog.

[→] Free scan

Run the seven steps on your own site.

One website, about 2 minutes. You get the full source list even if AI never names you.

  • Who AI names for your buying questions
  • Every page each answer cited
  • The pages to get listed on first
  • Free, no card
  • About 2 minutes
  • Every source linked

Want to look first? See how it works