How we measure AI visibility

The short version

We build a set of questions real customers ask about a trade in a specific area. We ask each question to several AI assistants, many times each. We count how often each business is named. Your result is a mention rate: the share of answers that named you.

1. Building the question set

A question set is the list of questions we ask for one trade in one area. Each set covers:

• The services people need. For roofing, that includes leak repair, storm damage, full replacement, and inspections.
• Urgency. "Emergency roof leak tonight" and "planning a roof replacement next spring" get different answers, so we ask both kinds.
• Different phrasings. People ask "best roofer near me," "who should I call for a roof leak in Tacoma," and "is Summit Ridge Roofing any good." We include direct requests, comparisons, and questions about price and trust.
• Location. Questions name the city or neighborhood, the way a customer would.

The question set stays fixed from month to month. If we changed the questions, a change in your numbers could just reflect different questions rather than a real change in your visibility.

2. Sampling the answers

Each question is asked to each assistant multiple times in fresh sessions. A single answer is one draw from a range of possible answers, and the businesses named can change between runs. Repeating the question shows the pattern underneath.

3. Counting a mention

A business counts as mentioned when an answer names it as a recommendation or option, whether or not a link is included. We match common variations of a business's name, such as "Rainier Roof" and "Rainier Roof & Gutter." A business that's only mentioned as one to avoid isn't counted as a recommendation, and we flag those answers separately.

4. The numbers in your report

• Mention rate: the share of answers that named you, overall and for each assistant.
• Share of recommendations: your mentions as a share of all business mentions in the answers. This shows how much of the conversation you own compared to competitors.
• Competitor rates: the same numbers for the businesses named most often.
• Sources: where assistants show their sources, the sites they cite most, such as review platforms and directories.

For small question sets, a rate can swing just by chance. When a change between months is too small to separate from that noise, the report says so instead of calling it a win or a loss.

5. How fixes are chosen

Fixes come from comparing you to the businesses that get named most. We look at the sources the assistants rely on, whether your business details are consistent across them, how clearly your website states your services and service area, and the volume and recency of your reviews. Each fix is ranked by how likely it is to matter and how much effort it takes.

What this can't tell you

• Exactly what any one person sees. Assistants personalize answers based on location, history, and settings. We measure a typical result, not every result.
• The future. Assistants update often. A strong result this month can change next month, which is why monitoring matters more than a one-time check.
• Why an assistant chose a business. We can see what assistants say and what they cite. We can't see inside them. Our fixes are based on patterns in what gets named, not on any inside knowledge.

No one can guarantee that an AI assistant will recommend a business. We measure, explain, and track. We don't promise placements.