Methodology
How we measure what AI says about you
No black box. Here is exactly which engines we ask, how often, how we find your business in the answers, and how the score is calculated.
1. Ask the questions customers ask
For each business we generate questions across six intents, in plain customer language, localized to the city. Market questions that don't name a business are shared by every business in that city and category, so everyone is measured against the same answers.
2. Ask every engine, more than once
AI answers vary from one run to the next, so we ask each question several times per engine (two runs by default, one for Google AI Overviews, more on higher plans) and on a schedule, rather than trusting a single screenshot.
3. Read the answers like a person would
We extract every business named in each answer, in order, plus the sources it cites. Then we match names to your business and competitors: exact names and aliases first, then your website or phone number in the answer, then fuzzy name similarity above a strict 0.6 threshold. Close calls are flagged as ambiguous, not counted.
4. Turn it into one number you can track
The AI Lead Score (0–100) combines five measurements. The weights are fixed and versioned, so a score from March is comparable with one from June.
AI Lead Score weights
Score version 1
- Mention rateHow often AI names you45%
- PositionHow high up when it does20%
- Engine coverageHow many engines know you15%
- Share of voiceYou vs. tracked competitors10%
- SentimentWhat AI says about you10%
An engine counts toward coverage when it names you in at least 25% of its answers. Sentiment comes only from questions that name you; a business AI never mentions gets no sentiment points.
Engines
Five AI engines, official access only
We use each provider's official API with web search turned on. We never scrape consumer chat apps.
ChatGPT
OpenAI API with web search
Perplexity
Perplexity API (search is built in)
Gemini
Google Gemini API with Search grounding
Claude
Anthropic API with web search
Google AI Overviews
Live Google results via a search-data provider
Questions
Six kinds of questions
Customers don't only ask for “the best”. We cover the ways people actually ask.
Discovery
“Who is the best plumber in Austin?”
Urgent
“I need an emergency plumber in Austin tonight”
Attribute
“Which Austin plumbers do tankless water heaters?”
Price
“Affordable plumber in Austin”
Comparison
“Rivera Plumbing vs. Radiant Plumbing”
Brand
“Is Rivera Plumbing any good?”
Limits
What our numbers can and can't tell you
- API answers closely approximate, but don't exactly match, what a signed-in person sees. Personal history, location and model updates all change answers.
- Scores are estimates from samples. That's why we repeat questions and track trends instead of reacting to a single run.
- Lead attribution relies on referrer data and your tracking script, so it counts AI-referred visits it can see. Some AI-influenced customers will always arrive without a trace.
- We don't control any AI platform, and nobody can guarantee a recommendation. Read our AI results disclaimer.