AI search visibility metrics should connect visibility to business outcomes, not stop at a count of mentions. For an established contractor, the useful scorecard has four layers: technical eligibility, appearances and citations, website engagement, and qualified jobs. When those layers are reviewed together, an owner can see whether the problem sits in discoverability, content, conversion, or lead handling.
I would not manage AI visibility from one vendor score. Google, ChatGPT, and other answer systems expose different data, and a homeowner may see a company in an answer before returning later through branded search, Maps, or a direct call. A practical scorecard needs to respect those limits while still giving the team something concrete to improve.
Which AI search visibility metrics belong on the scorecard?
The four layers are eligibility, visibility, engagement, and business outcomes. Each answers a different question, and no single layer proves that AI search is producing revenue.
- Eligibility: Can search and AI crawlers access, understand, and cite the important pages?
- Visibility: Does the company appear or earn citations for a consistent set of valuable questions?
- Engagement: Do people visit useful service and location pages after encountering the brand?
- Outcomes: Do those visits and assisted journeys produce qualified calls, appointments, booked jobs, and sold work?
This structure keeps an early signal from being mistaken for a result. A citation can matter without generating an immediate click, while a referral visit can look encouraging and still produce an irrelevant lead.
Layer one: confirm that the business is eligible to appear
Eligibility is a prerequisite, not a performance win. Google says pages need to be indexed and eligible to appear with a snippet to be included as supporting links in its AI features. Its current guidance also says there is no special AI schema or machine-readable file required for inclusion.
For ChatGPT search, OpenAI says publishers should allow OAI-SearchBot if they want public pages to be discoverable and cited. That is separate from the controls used for model training. A contractor should therefore check crawler access, indexability, canonical URLs, and whether key service and location pages contain accurate public business details.
The useful eligibility line on a monthly scorecard is simple: the percentage of priority pages that are indexable, crawlable by the relevant search bots, and free of obvious technical conflicts. Google provides more detail in its guidance for appearing in AI search features, while OpenAI explains its crawler and referral behavior in the publisher and developer FAQ.
Layer two: measure appearances with a fixed question set
AI visibility should be sampled against the same high-value questions over time. A plumbing company might track emergency water-heater questions in each real service area. An HVAC company might track replacement, repair, financing, and maintenance questions that match the work it wants.
I recommend a small, stable question set grouped by service, location, and homeowner intent. Record whether the company is named, whether its website is cited, which page receives the citation, and which other sources appear. Keep the test conditions consistent enough to reveal direction, then treat the results as a sample rather than a census.
Generated answers can vary by wording, location, personalization, product, and time. A gain in sampled mentions is evidence of improving visibility, but it is not a guaranteed market-share calculation. That distinction matters when a dashboard assigns a precise-looking percentage to a limited set of questions.
Google added dedicated generative-AI performance reporting to Search Console for a subset of websites in June 2026. When available, the report can show impressions, clicks, pages, countries, devices, and dates connected with Google’s generative AI experiences. Google’s announcement of the generative AI performance report also makes clear that availability is still limited, so contractors should not assume every property has it.
For a broader view of the factors that can support local discoverability, I have also broken down AI search ranking factors for local businesses. That is useful context, but the scorecard here is meant to measure movement and business value rather than produce another optimization checklist.
Layer three: separate referral visits from total influence
Referral traffic is measurable, but it captures only part of AI search influence. OpenAI says ChatGPT adds utm_source=chatgpt.com to referral URLs, which allows those sessions to be identified in analytics. Google Analytics can then report the session source and medium in its Traffic acquisition report.
A monthly view should include AI referral sessions, the landing pages receiving them, engaged visits, and conversion actions. Google’s Traffic acquisition documentation explains how session source and medium describe where visitors came from.
The caveat is attribution. A homeowner may read an AI answer, remember a company name, search for it later, open its Business Profile, and call from Maps. That journey may never appear as an AI referral. Referral sessions are therefore a verified subset of influence, not the total effect.
If the technical setup or discoverability picture is unclear, Revved’s AI and local visibility audit can identify the gaps worth checking first.
Layer four: connect attention to qualified jobs
Business outcomes are the deciding layer. The scorecard should connect landing pages and known sources to qualified calls, form submissions, appointments, booked jobs, and sold revenue wherever the company’s tracking systems support that connection.
For trades businesses, raw lead counts hide too much. A restoration inquiry outside the service area, an HVAC job the company does not offer, and a booked replacement estimate should not carry equal weight. The office needs a consistent qualification status and a reliable handoff from call tracking or forms into the CRM.
Use the gaps between layers to choose the next action
The pattern between layers is more useful than any isolated total. If priority pages are blocked or missing from the index, fix eligibility first. If pages are eligible but absent from relevant answers, improve the underlying service evidence, local specificity, and usefulness.
If mentions and citations rise without referral visits, inspect which pages are cited and whether the answer gives searchers a reason to continue. If visits rise without qualified inquiries, the likely issue shifts toward intent, page clarity, trust, or conversion. If qualified inquiries arrive but bookings do not, the website may have done its job and the handoff or call process needs attention.
This diagnostic approach prevents a common mistake: asking the content team to publish more when the real constraint sits in tracking, dispatch, service coverage, or follow-up.
A monthly review should end with one accountable decision
AI search visibility metrics are useful when they lead to a specific operational choice. Each monthly review should name the weakest layer, the evidence behind that judgment, one action, one owner, and the date the result will be checked.
For one contractor, that may mean opening crawler access to an important service page. For another, it may mean adding real project proof to a location page, fixing call qualification, or improving the page already earning citations. The right action depends on where the scorecard shows the chain breaking.
Established trades companies that need the website, local presence, acquisition, and follow-up systems to work as one can explore Revved’s connected implementation program. The goal is a reporting system that shows what deserves attention next, then follows that work through to booked jobs.