SEO & GEO

Google AI Impressions: What to Check in Search Console

Read Google’s AI impressions without mistaking them for clicks or inquiries.

What the report can tell you

A business owner sees more appearances in Google's AI features, but the inbox is as quiet as before. Nothing in that observation is contradictory. A page can appear as a supporting link without receiving a visit. A visit can happen without producing a useful inquiry. Before changing the whole site, find out which of those steps you can actually observe and which step is worth improving.

Google says it completed the worldwide rollout of the dedicated generative AI insights on August 31, 2026. In Search Console, the Search report covers impressions from AI Overviews and AI Mode. Discover has a separate generative AI report; it is not a tab for the same audience, and it is not a proxy for search demand. Search Labs experiments are excluded. Worldwide rollout does not mean every property will show both reports. Google says insufficient impressions are a common reason for a missing report. Eligibility settings can matter as well.

If the Search report is missing, first confirm which property you opened: a URL-prefix property can exclude pages on a different host or protocol. Check one important URL with URL Inspection and review the site’s Search generative AI settings if someone changed them. An indexed page is eligible for consideration only if it also satisfies the relevant display requirements; inspection does not prove that the page appeared in an AI answer. Do not widen access settings merely to make a report show up.

Separate visibility, traffic and inquiries

Visibility: Use the dedicated AI report to see whether links to your pages appeared in supported features. An impression is not a visit, a unique person or proof that someone read the answer or noticed your brand. The report does not show the exact answer in which each link appeared, and it is not a measure of your visibility in other AI products.

Traffic: Use the standard Search performance report, with the Web search type, to see Google Search clicks for a page. It includes activity from Google's AI Search features alongside other Web activity. The dedicated report is a view into impressions already included in that data, not a new quantity to add to it. The dedicated view does not break out AI clicks. Dividing total Web clicks by AI impressions would mix two different groups and would not yield an AI click-through rate.

Inquiries: Count confirmed form submissions, booked appointments or qualified contacts in your own analytics or CRM. A tap on a phone or messaging button can be useful as an intermediate event, but it is not the same as a call answered or a request received. Check whether your team can reconcile the event with an actual inquiry. Do not describe a change in those totals as coming from Google AI unless you have reliable evidence that supports that attribution.

Set up a baseline you can repeat

  • Choose the verified Search Console property that contains your canonical URLs. Open the dedicated Search report, select a complete 28-day period and note the exact dates. Avoid drawing conclusions from the newest preliminary data. If you compare with your analytics, remember that the report uses Pacific Time; your business dashboard may use local dates.
  • Start in Pages. Choose one URL that serves a business goal, such as a service page that explains what customers can order. Record its AI impressions. Write down filters for device, country and search type; if you use the report's text or multimodal filter, reuse it next time. When a URL seems to be missing, check the canonical version rather than assuming that Google ignored the page.
  • Open the standard Search results report and record Web clicks for the same page and dates. Separately, log the qualified inquiries you can associate with that landing page and write down how you counted them. Leave gaps marked as unavailable. In exports, a zero may represent a value unavailable in the interface, so preserve that distinction in your notes.
  • Compare with the previous complete period. Keep the windows the same length and change one filter at a time. The chart can count by property while the Pages table counts by URL. If two URLs from the same site appeared together, summing page rows can give a different result from the chart. The table also has row limits, so do not use its sum as a replacement for the chart total.

Start with a question, not an alert: is visibility changing for a page that matters, and can a buyer take the next step there? Site-wide movement may be driven by a few unrelated informational pages. The page-level comparison helps you avoid prioritizing whichever number happened to move the most.

A small-business example: branded workwear

Imagine a workwear supplier whose service page appears in the AI report. This is a hypothetical example, not a client result. The page shows examples of embroidery but does not say whether a small team can order a short run. On a phone, the visitor has to open a generic contact form before learning what information the supplier needs. Ask the sales team what it actually checks before quoting: garment, decoration method, quantity and deadline.

Add a short answer beside the quote button. If the minimum varies, explain which choices determine it and let visitors request confirmation. Ask for only the details needed to respond, and submit a test request on a phone. Note the edit date and what changed. The point is to reduce uncertainty for someone already looking at the offer, not to add SEO text or promise that Google will show the new paragraph.

Read the follow-up without overstating it

After another complete 28-day period, repeat the comparison. More AI impressions alongside unchanged Web clicks indicate greater reported visibility without a corresponding increase in aggregate clicks. They do not tell you whether AI clicks rose or fell. More qualified inquiries with flat impressions could still be good for the business; check whether the change came from the page, another channel or how the team records requests. If all three move together, that is a reason to investigate, not proof of cause.

Check promotions, seasonal demand and reporting anomalies before explaining a swing. If traffic is low, wait longer rather than treating a handful of events as a trend. If there is no dedicated report, use the standard Search report and the page's inquiry data to choose a useful improvement. Keep the minimum next step simple: pick one page, answer one real buying question, and set a date to review the same measurements.

Read the original source