Audit a contractor’s visibility in Google AI Overviews and ChatGPT with two separate measurements. Use Google’s generative AI performance report in Search Console to find pages earning AI impressions. Then run a fixed set of local buying questions in ChatGPT search and record whether the company is named, whether its website is cited, which competitors appear, and whether the answer fits the requested service and market.
Do not collapse those results into one vague “AI visibility score.” Google now supplies first-party impression data. ChatGPT requires a controlled observation log. The useful output is a list of missing queries, weak pages, conflicting business facts, blocked crawlers, and lead paths that need repair.
What an AI visibility audit can prove
A defensible audit answers four different questions:
- Eligibility: Can Google and OpenAI access the pages that should represent the business?
- Appearance: Did a page receive a Google generative AI impression, or did the contractor appear in a tested ChatGPT answer?
- Attribution: Was the contractor merely named, or was its own website cited as a source?
- Commercial outcome: Did the visit produce a call, form, booking, or qualified conversation?
These are not interchangeable. A citation without a good service page can send the wrong visitor. A mention without a link may create awareness but no measurable website session. A Search Console impression shows appearance in Google’s generative AI features, not a booked job.
Start with Google’s first-party AI visibility report
Google announced dedicated Search Console reports for generative AI features, including AI Overviews and AI Mode. Google says the reports provide dedicated views of impressions within those features.[1] Its documentation says the insights had rolled out to all websites worldwide by August 31, 2026.[1]
The Search report can show changes in organic generative AI impressions, the pages with the highest or lowest impressions, and where those impressions originate by device or country.[4] That gives a contractor a first-party baseline that did not exist in the same form earlier in 2026.
Export the baseline before changing the site
- Open the verified domain property in Google Search Console.
- Open the generative AI performance report for Search.
- Set a fixed comparison period, such as the most recent 28 complete days and the prior 28 days.
- Export the available page, country, device, and date data.
- Mark each page by trade, service, market, and search intent.
- Save the export date and property name with the file.
Do not begin by chasing the page with the largest number. First ask whether the visible pages match profitable work. An HVAC company may earn impressions for a thermostat troubleshooting article while its replacement, maintenance-plan, and emergency service pages remain absent. The report shows where Google used the site. It does not decide whether that visibility is commercially useful.
If the report is absent or empty, record that result. Google’s help page notes that a property may not show the report when the site has not received enough impressions in generative AI features.[4] An empty report is not proof that the business never appears anywhere in AI search.
Audit ChatGPT with a controlled query basket
ChatGPT does not give a contractor the same Search Console report described above. Use a repeatable observation process rather than a single question typed once.
Create 15 to 25 questions based on jobs the contractor wants. Keep the wording, location, platform, search setting, and test date. Split the questions by intent:
| Query group | Contractor example | What to record |
|---|---|---|
| Urgent problem | Who repairs a leaking water heater near [city]? | Businesses named, service match, location match, sources |
| Planned replacement | Which local HVAC companies install cold-climate heat pumps? | Named contractors, stated qualifications, cited pages |
| Trust and verification | How can I verify a roofer’s license in [state]? | Regulators cited, contractors mentioned, accuracy |
| Comparison | Compare local electricians for a 200-amp panel upgrade in [city] | Shortlist, comparison basis, missing proof |
| Service specificity | Find a remodeler near [city] with aging-in-place bathroom experience | Project evidence, service-area evidence, unsupported claims |
Use realistic questions, not repeated variations of the company name. A branded question tests whether the system can find a known entity. A non-branded buying question tests whether the company enters the consideration set.
Record the answer, not your impression of it
For every test, capture:
- exact question and location wording;
- date, platform, account state, and whether web search was active;
- contractors named and their order of appearance;
- source links shown in the answer;
- whether the contractor’s own domain was cited;
- the service, geography, license, review, or project claim attached to the company;
- any factual conflict or unsupported statement.
Repeat the same basket at a later checkpoint. Generated answers can change, so one favorable response is a signal rather than a stable rank. Do not report “number one in ChatGPT” from one answer.
Check whether the site can enter ChatGPT search
OpenAI identifies OAI-SearchBot as the crawler used to surface websites in ChatGPT search. Its documentation says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, although they may still appear as navigational links.[5]
Check the live robots.txt file, page-level robots directives, canonical tags, status codes, and any firewall rules that treat crawler requests differently from ordinary visitors. Test the important service and location pages, not only the homepage.
Access is only the first gate. A crawlable page can still be too vague to support a local recommendation. It needs a clear service, service area, business identity, useful evidence, and a working route to contact the contractor.
Inspect the pages that win and the pages that disappear
Match every Google impression and ChatGPT citation to the exact source page. Then review the page as a customer and as a record.
Page evidence checklist
- The page names one real service or problem in plain language.
- The service area is supported by operating facts, not a swapped city name.
- The public brand, legal or licensed relationship, address type, and phone path agree.
- Claims about licenses, certifications, warranties, equipment, and financing can be verified.
- Project examples, photos, process details, and customer questions add information beyond generic advice.
- The title, main heading, canonical URL, internal links, and structured data describe the same page.
- The call, form, or booking path works on a phone.
Use the contractor entity trust checklist when names, licenses, addresses, or phone numbers conflict. Read the contractor schema guide when visible facts and structured data disagree. Schema can clarify a real business. It cannot supply proof the page does not contain.
Google’s current guidance says no special optimization is required to appear in AI Overviews or AI Mode.[3] Google treats optimization for generative AI search as part of search optimization and recommends useful, original content for the audience.[2] That is a useful boundary. Do not buy an “AI trick” while the site has duplicate pages, inaccessible content, weak local proof, or dead lead forms.
Compare competitors at the page and evidence level
A contractor-owner needs to know why another company appeared. Record the cited competitor page and the evidence it carries. Look for a better service match, clearer city coverage, stronger project documentation, more coherent business identity, or a source outside the contractor’s own website.
Do not copy a competitor’s claim merely because an answer repeated it. Verify licenses, certifications, office locations, and other consequential facts at their primary records. Separate what the company claims from what the record demonstrates.
The related guide, why AI search leaves local contractors out, covers the underlying eligibility and evidence failures. The broader contractor AEO operating guide connects those findings to a 90-day repair cycle.
Use a result sheet that leads to action
Keep platform evidence separate, then summarize the work in one operating sheet.
| Finding | Evidence | Decision |
|---|---|---|
| Google AI impressions exist, but only for informational articles | Search Console page export | Strengthen internal paths and missing service pages |
| ChatGPT names the company but cites another domain | Saved answer and source list | Find the outside proof and improve the matching owned page |
| Competitors appear for a profitable service the site barely explains | Fixed query basket and page comparison | Build or repair the service evidence |
| OAI-SearchBot is blocked | Live crawler and robots test | Confirm policy, then correct access if authorized |
| Visibility produces visits but no leads | Analytics, call, form, and booking records | Repair the offer and conversion path |
Retest after material changes using the same question basket and comparable date window. Change one major variable at a time when possible. That makes the next result useful instead of merely different.
Turn the visibility gap into a repair decision
The free KKS contractor audit includes a manually reviewed 15-factor scorecard, comparison with up to three relevant local competitors, search and demand evidence when available, public license and ownership facts when records match, and a sequenced 90-day plan. The recommendation may be a focused repair, ongoing management, a managed rebuild, or no engagement. Qualified reports are delivered within two business days after KKS confirms the business, market, and evidence scope. No credit card or required sales call.
Does weak AI visibility require a new website?
No. A blocked crawler, incorrect canonical, missing entity explanation, weak internal link, or thin service section may be repairable. A rebuild becomes reasonable when the owner cannot control the domain or hosting, the system keeps generating duplicate pages, important content cannot be edited cleanly, or the lead path cannot be tracked and maintained.
Start with evidence. The audit may support a technical repair, a content change, an existing-site management plan, a rebuild proposal, or no engagement. Read the contractor SEO strategy before separating AI visibility from the rest of search and conversion work.
Frequently asked questions
Can Search Console show Google AI Overview visibility?
Yes. Google’s generative AI performance report shows impressions from generative AI features on Search and identifies visible pages plus available device, country, and date data. Treat an impression as appearance evidence, not as a lead or booked job.
Can a contractor check its ChatGPT ranking?
There is no single stable contractor ranking in the sources reviewed. Use a fixed basket of local buying questions, save the answer and cited sources, and retest under recorded conditions. Report mentions and citations rather than claiming a permanent position.
How many AI search questions should a contractor test?
KKS recommends 15 to 25 questions for a focused manual baseline. Cover urgent problems, planned work, comparisons, trust checks, and service-specific needs. Keep the same core basket for later comparisons.
Does blocking GPTBot keep a contractor out of ChatGPT search?
Not by itself. OpenAI documents GPTBot and OAI-SearchBot as separate controls. OAI-SearchBot is the crawler tied to inclusion in ChatGPT search answers.
How often should AI visibility be audited?
Capture a baseline before major changes, then retest after the pages have been crawled and enough comparable data exists. A monthly operating check is useful during active repairs. Stable sites can use quarterly reviews plus checks after a move, rebrand, new service, or major platform change.
Sources
- Introducing Search Generative AI performance reports in Search Console. Accessed September 5, 2026.
- Google's guide to optimizing for generative AI features on Google Search. Accessed September 5, 2026.
- AI features and your website. Accessed September 5, 2026.
- Generative AI performance report (Search). Accessed September 5, 2026.
- Overview of OpenAI crawlers. Accessed September 5, 2026.
Platform and documentation snapshot: September 5, 2026. Search features, reports, crawler controls, and generated answers can change. Verify the live documentation and record the test conditions used for each audit.