How LeadValets Tests Search and AI Visibility

We test visibility using the questions that real buyers type into Google and AI answer engines — not generic keyword lists. Queries are built from four components: the trade (e.g., "plumber"), the service (e.g., "water heater repair"), the city or service area, and the intent type (emergency, high-value commercial, best-company comparison, general local service, or informational).

Each tracked project uses a set of rendered questions generated from PromptTemplate records, with variables filled in from the client's BusinessLocation data. The same question set is sent to every AI engine so results are directly comparable.

AI answer engines and Google results both vary by the searcher's location. To produce consistent, comparable measurements, every query is run with an explicit geographic context tied to the client's service area.

We test the AI answer engines that plumbing and HVAC buyers are most likely to use when researching a contractor. Each engine is tested with the same rendered question, and the full response is saved.

For every response, we record:

• Date — the timestamp the response was collected.
• Engine — the AI answer engine and model/surface version (e.g., ChatGPT GPT-5, Gemini 3, Perplexity Pro, Claude Sonnet, Google AI Overviews).
• Prompt — the exact rendered question text sent to the engine.
• Full response — the complete text returned, stored verbatim in the ProviderRun record.
• Source URL and screenshot — when available, the URL and a dated screenshot of the response.

These three terms describe different levels of visibility in an AI answer. They are not interchangeable — a business can be mentioned without being recommended, and cited without being mentioned by name.

The Tracker measures each of these as rates: mention rate, recommendation rate, primary recommendation rate, and citation rate — each calculated as the percentage of eligible responses where the event occurred.

Each tracked project includes a set of competitor businesses so we can measure share of voice — not just whether the client is visible, but how visible they are relative to the other companies a buyer might choose.

Case study results are reported as before-and-after comparisons with dated evidence. We do not report a ranking change without a dated screenshot or a call-tracking log entry that shows the starting and ending state.

The Search Authority Snapshot runs a server-side crawl of your homepage — the same way a non-JavaScript crawler would see it. Each check is binary (pass or fail), and the two scores are the sum of weighted passes. No check is ever staged or simulated — if we cannot fetch your site, we tell you.

Full Audit: AEO Score (0–100)

The full Search Authority Audit™ produces an overall AEO score from seven weighted categories. Each category is scored 0–100 independently, then combined using the weights below.

We are transparent about what our data can and cannot tell you. AI answer engines are non-deterministic — the same prompt can return different responses on different runs. We account for this by testing the same question set repeatedly and reporting rates, not single observations.

If you find an error in any LeadValets report, case study, or methodology description, we want to know. Contact us at jfloyd@leadvalets.com with a description of the error and, if applicable, a link to the page or report.

When a correction is made, the "Last updated" date at the top of this page is updated to reflect the change.