Measure AI search in three separate layers: where your business appears, whether people reach you, and whether those inquiries become suitable projects. A generated answer mentioning your company is an observation. An attended consultation with a source recorded is a business outcome. Keep both, but do not report them as the same metric.
Start with a monthly worksheet your office can actually maintain. A short, consistent record is more useful than a dashboard full of unverifiable scores. The goal is to learn which information and inquiry paths deserve work.
Use Google's current report for Google visibility
Search Console's Generative AI performance report covers impressions in AI Overviews and AI Mode. Google states that the insights rolled out worldwide by August 31, 2026; a property may lack the report when it has insufficient impressions. The documentation describes page, country, date, and device views, not contractor revenue.
Keep a repeatable set of discovery questions
Write a question set around actual work: a service-and-town request, a project constraint, a cost-planning question, and a provider comparison. Include both branded and unbranded questions. Save the wording and location context so next month's checks have a useful reference point.
For each observation, record the platform, date, question, business names shown, cited pages, and any inaccurate company details. If you run the same question again and get a different answer, preserve that difference. Do not average a handful of manual checks into a claim about your market share.
- Discovery question: who builds custom pools with small-yard access in our service area?
- Planning question: what affects the schedule for a lived-in kitchen remodel?
- Brand question: what services does our company provide and where?
- Comparison question: what should a homeowner compare before choosing a window installer?
Record referrals and homeowner explanations
Where your analytics captures a referring website, keep that signal. Also give the office a simple first-touch question: how did you first hear about us? Save the answer in the homeowner's own words before assigning a standardized source category.
A prospect might discover you through an AI answer, check your reviews, and return by searching your name. Last-touch analytics can then show branded search rather than the earlier AI interaction. Preserve both facts when available. If the source is unknown, record unknown rather than guessing that a rise in direct visits came from AI.
Build a dashboard with honest denominators
Use separate rows for Google AI impressions, observed mentions, visits with an identifiable AI referrer, self-reported AI inquiries, qualified inquiries, attended appointments, and signed jobs. State the reporting window. Do not divide all inquiries by impressions from only one platform to create an all-platform conversion rate.
Illustrative example: the office records six prospects saying they used AI, four meet your service criteria, three attend, and one signs. Report those counts with the period and the incomplete-attribution caveat. It is one observed path in that business, not an expected conversion rate for contractors.
Turn the findings into a page or process change
If repeated answers misunderstand your territory, correct the website and business profiles. If a cited page answers the wrong kind of project, refine its scope and connect it to the appropriate service. If visits arrive but inquiries fail, test the mobile form and clarify what the first consultation includes.
Choose one change with an owner, publish it, and annotate the date. Review the subsequent period while considering other campaigns, seasonality, and operational changes. You can identify a useful pattern without claiming that a before-and-after comparison isolates one cause.