Google has rolled out dedicated generative AI performance reports in Search Console to websites worldwide. For the first time, a verified site owner can see which pages are receiving impressions in AI Overviews and AI Mode without relying on occasional screenshots or third-party estimates.
That is useful, but it is easy to read too much into the new numbers.
An AI impression does not prove that somebody visited the website, requested a quote or became a customer. The report also does not reveal the customer’s exact query. For builders, plumbers, electricians, roofers and renovation companies, the sensible goal is therefore not “get the biggest AI number”. It is to learn which real services, projects and advice pages Google is using, then test whether that visibility supports better enquiries.
This guide explains what the report can show, what it leaves out and how to turn it into a practical monthly decision.
What Google has launched
Google announced the generative AI performance reports in June 2026 and later added that the insights had been rolled out to all websites worldwide by 31 August.
There are separate views for Search and Discover. This article focuses on the Search report, which includes impressions from AI Overviews and AI Mode. Google’s report documentation says Search Labs experiments are not included.
This is a reporting release, not a confirmed ranking update and not a new optimisation requirement. A site does not receive an advantage simply because somebody opens the report. It provides evidence about visibility that was previously difficult to separate.
Google’s broader guidance for generative AI features still points businesses towards accessible pages, useful original content and established SEO practices. The report helps judge that work; it does not replace it.
Understand what an AI impression means
In this report, an impression is counted when a link to the website is shown to a user in a supported generative AI feature. If two results from the same property appear in one AI result, Google says the property-level chart counts a single impression. Adding a URL filter changes the chart to URL-level aggregation.
That definition matters. It means the top-line number is a measure of appearances, not individual links, visits or prospective customers.
A builder’s extension guide could receive an impression when it is cited in an AI Overview. The user may follow the link, continue reading the generated answer or choose another source. Search Console’s AI report records the appearance, but it cannot tell the business which of those outcomes occurred.
Treat an impression as the start of an investigation:
- Which page appeared?
- What job does that page perform for a customer?
- Does the page lead naturally towards a relevant service?
- Did normal organic visits and qualified enquiries move in the same period?
That keeps the new metric in proportion.
Read the available dimensions carefully
The Search report currently provides impressions with four useful dimensions. Each answers a narrow question.
| Dimension | What it can tell you | What it cannot prove |
|---|---|---|
| Pages | Which final, usually canonical URLs appeared | The precise prompt or why Google selected the page |
| Countries | Where the searches originated | The customer’s town, service area or commercial value |
| Devices | Whether the appearance was on mobile, desktop or tablet | Whether the user contacted the business |
| Dates | How impressions changed by day, week or month | That a content edit caused the change |
Google assigns most data to the canonical URL and groups the page dimension by the final URL after redirects. That can explain why an old address does not appear separately after a clean redirect.
The date view uses Pacific Time rather than UK local time. A late-evening impression in Britain may therefore sit under a different date from an internal enquiry record. For monthly decisions this is rarely a serious problem, but it makes day-by-day matching unreliable.
The report does not currently provide a query dimension, clicks, click-through rate or average position. Do not reverse-engineer a supposed prompt from a page title and present it as fact.
Establish a baseline before improving anything
Start with a consistent comparison rather than reacting to the latest point on the graph.
Open the Search Console property that represents the complete site. Choose a full recent period, export the chart and page table, and record the date range. For a small trade website, monthly data will usually be easier to interpret than daily movement because the totals may be low.
Record five things:
- Total generative AI impressions for the period.
- The pages with the most impressions.
- Pages that appeared for the first time.
- The main countries and device split.
- Significant website or service changes made during the period.
The newest figures can be preliminary and may change while Google finishes processing them. The standard report limitations also apply, including limits on table rows. Preserve the export rather than assuming today’s screen will remain identical.
If Search Console access or indexing is uncertain, resolve that first. A focused SEO audit for the website can check whether important pages are crawlable, canonical and internally linked before anybody interprets their absence from an AI report.
Classify visible pages by their customer role
A list of URLs becomes more useful when each page has a commercial role. I recommend assigning every visible page to one of five groups.
Service pages
These explain work the business actually sells: rewiring, roof replacement, extensions, bathroom renovation or commercial plumbing. An AI impression can indicate that Google found the page relevant enough to surface, but the page still needs to help a suitable customer decide what to do next.
Check whether the offer, coverage, process, evidence and next step are clear. If several related services are squeezed into one vague page, service-page planning may be more valuable than adding another blog post.
Project and case-study pages
These can provide first-hand details that generic publishers do not have: property type, access constraints, material decisions, work sequence and the finished result. The guide to turning construction projects into useful case studies shows how to capture that evidence without inventing outcomes or exposing customer information.
If one case study earns AI visibility, connect it clearly to the service it demonstrates. Do not leave the reader at the end of a gallery with no route to understand the offer.
Advice articles
Guides often answer early research questions. Their value is not only direct quote requests. A useful article can introduce the business, demonstrate experience and lead a reader towards a service when the connection is genuine.
Review whether the advice is current, specific and supported. Ongoing content optimisation should strengthen a valid page before the business creates five overlapping answers to the same question.
Location pages
Location pages should explain a real service-area difference, not swap one town name for another. AI impressions do not make a thin doorway page useful. A strong local page still needs distinct customer information and proof connected to the area.
The homepage
Homepage visibility may relate to the brand, the overall offer or a broad business question. Check that visitors can immediately understand what the company does, where it works and how to enquire.
Connect AI visibility with visits and qualified enquiries
The AI report is deliberately narrower than the ordinary Search performance report. Google’s standard performance documentation explains that the main report can show clicks, impressions, click-through rate and average position, with dimensions such as queries and pages.
Use the reports together, but do not pretend they provide perfect attribution.
For each AI-visible page, compare three layers:
- Visibility: generative AI impressions in Search Console.
- Engagement: organic landing-page visits and useful actions in website analytics.
- Commercial outcome: calls, forms or messages that become relevant site visits, quotations or won work.
A defined lead-tracking process is the third layer. Without it, a business can report impressive visibility while having no idea whether the enquiries suit its services, value or working area.
Suppose a roofing maintenance guide gains AI impressions. If organic visits and qualified roof-inspection enquiries also rise, the page deserves closer attention and perhaps a stronger connection to the maintenance service. If impressions rise while visits and leads remain flat, check the page’s intent and enquiry route before declaring success or failure. The AI appearance may be informational, the sample may be small, or the on-page next step may be weak.
This is correlation, not proof that an AI feature caused every visit. Use the evidence to choose a measured test, not to manufacture an attribution claim.
Improve the page that earned the evidence
When a page appears in the report, resist the urge to create a separate “AI version”. Improve the useful asset that already exists.
Depending on its role, that may mean:
- adding a missing project constraint or decision
- replacing generic claims with permission-backed photographs and real process detail
- clarifying the service and working area
- answering the next question a suitable customer genuinely asks
- linking the advice page to the relevant commercial service
- making the mobile enquiry route easier to use
- updating an obsolete fact and showing the source near the claim
Our guide to Google AI search optimisation for builders explains why Google does not require special AI files, tiny answer fragments or a page for every prompt variation. Useful first-hand information remains the advantage.
Technical quality still matters too. A slow, unstable or difficult mobile page can waste the attention it earns. The Core Web Vitals guide for construction websites helps separate loading, responsiveness and layout-shift problems without treating a single lab score as the whole customer experience.
Use a simple 30-day review cycle
A monthly routine is enough for many small construction and home-improvement businesses.
Day 1: export and label
Export the AI page data, label each URL by customer role and compare it with the previous complete period. Note launches, major edits and technical changes.
Week 1: choose one page
Select one page where improvement has a clear business reason. A service page with impressions but an unclear quotation route is usually a better priority than an unrelated article with the largest number.
Week 2: make one controlled improvement
Add evidence, improve the explanation, repair internal links or clarify the call to action. Avoid changing the title, page content, URL and tracking at the same time because the result becomes harder to interpret.
Week 3: connect the journey
Link the improved page from a relevant service, project or guide. Where local intent matters, align it with accurate local SEO foundations rather than forcing town names into the copy.
Day 30: compare all three layers
Review AI impressions, ordinary organic engagement and qualified enquiries. Keep the change if it improves the customer journey even when AI impressions remain flat. The purpose is better business evidence, not pleasing one chart.
If the report is missing or empty
Google says the insights are available worldwide, but its help page also lists two reasons a property may not show the report: access may still be rolling out for that property, or the site may not have received enough supported impressions. A site can also make itself ineligible for generative AI features through certain preview controls.
Check the correct verified property, user permissions and any relevant Search preview directives. Then wait for a meaningful reporting period. Do not state that the website has “zero AI visibility” merely because a dedicated report is absent.
The right fallback is ordinary SEO work: make important pages accessible, useful and connected. The September SEO watchlist for builders and trades provides a broader measurement plan covering confirmed updates, local visibility, website performance and enquiry routes.
Final thoughts
Search Console’s generative AI report replaces guesswork with a limited but valuable piece of first-party evidence. It shows appearances in AI Overviews and AI Mode by page, country, device and date. It does not show the exact query, clicks or revenue.
Video needs the same measurement discipline. The practical guide to video SEO for builders and trades separates video indexing and impressions from watching behaviour and qualified enquiries.
For builders and trades, the useful question is not “How do we maximise AI impressions?” It is “Which of our real pages are being surfaced, and how can they better help the right customer?”
Export a baseline, classify the visible pages, connect them to normal engagement and qualified leads, then improve one genuine asset at a time. That produces better decisions whether AI visibility rises quickly, slowly or not at all.
Frequently Asked Questions
Why can’t I see the generative AI performance report?
Google says a property may not show the report if rollout access is not yet available for it or if the site has not received enough impressions in supported generative AI features. Check the correct property and permissions, but do not treat an absent report as proof of zero visibility.
Does the report show the searches or prompts people used?
No. The current report provides page, country, device and date dimensions for impressions. It does not include a query or prompt dimension.
Does an AI impression mean somebody visited the website?
No. It means a link from the site was shown in a supported generative AI Search feature. Use ordinary Search performance, analytics and qualified-enquiry records to assess what happened beyond the appearance.
Should a builder create separate pages for AI search?
Not simply because the report exists. Improve useful service, project and advice pages for real customers. Google’s current guidance says established SEO practices and valuable original content still apply to its generative AI features.
