Discussion
  • AI Search
  • Search Console

Search Console’s generative AI report shows impressions but no clicks. What are you doing with that data beyond reporting it, and has it changed any decision?

Published 1 reply

Discussion starter: an AI search question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.

Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.

1 reply

  • admin
    Admin
    Original poster

    Taken on its own, an impressions-only report cannot tell you what AI exposure is worth, so reporting it as a headline number mostly produces a chart nobody can act on. It becomes useful once you join it to page-level clicks and conversions and use it to sort pages into groups that need different decisions. Since 31 August 2026 it has one decision it clearly should inform: whether to use Google’s new sitewide control for opting out of generative AI features, because this report is the only first-party measure of what that switch would remove.

    On “has it changed any decision”: the report reached all sites worldwide less than a month ago (the UK had it from June), so nobody has a year of data behind it and public evidence of outcomes is thin. The decisions below are the ones it has a legitimate claim to change, along with the evidence that should trigger each one.

    What the report can and cannot tell you

    According to Google’s help page, the report counts how many times links to your site were shown in a generative AI feature on Google Search (AI Overviews and AI Mode), excluding Search Labs experiments. You can break it down by page (grouped by canonical URL), country, device, date and search type (“Web: text-based” and “Web: multimodal”), and export the chart and table. It draws on the Web search type data and carries the usual 1,000-row limit. Google’s announcement also covers generative AI features in Discover.

    What it lacks shapes everything you can do with it: no queries, no clicks, no CTR and no position. And because AI Mode clicks are counted in Web totals and, according to practitioner reports, arrive in GA4 as google / organic, you cannot recover AI clicks from analytics either.

    Watch for this:

    Do not subtract AI impressions from Web impressions to estimate “classic” impressions. Google does not document how a URL that appears in both an AI feature and the standard results on the same page view is counted. Use the ratio of AI impressions to Web impressions to rank pages against each other, not as a literal share.

    Turning it into a page-level segmentation

    The most useful step is also the simplest. Export the generative AI report by page for two comparable periods. Export Web performance by page for the same periods, and GA4 conversions by organic landing page. Join the three on URL, then group pages by template or section (guides, category pages, product pages, tools, location pages), because single-URL numbers are noisy.

    Plot each page group on two axes: how much AI exposure it has relative to its Web impressions, and how its organic clicks changed. Four groups fall out, and each points to a different decision.

    High AI exposure, clicks falling

    The at-risk set. Check whether conversions fell with clicks. If they held, the lost clicks were probably low-intent. If they fell, this is where content and product investment should go first.

    High AI exposure, clicks stable

    Being cited has not cost you traffic here, at least not yet. Protect these pages: keep them accurate and current, and resist restructuring them for “AI-friendliness”.

    Low AI exposure, clicks falling

    The decline is probably not about AI features. Look at rankings, competitors, SERP layout, seasonality and technical issues before blaming AI Overviews.

    Low AI exposure, clicks stable

    Your baseline, and a natural control group when you want to judge whether a change elsewhere is AI-related.

    The third quadrant is the one that stops bad decisions. Without this report, teams tend to attribute any organic decline to AI Overviews. Pages with falling clicks and little AI exposure are direct evidence that something else is going on.

    Hypothetical example:

    A B2B software site has three main page groups. Glossary pages carry heavy AI exposure and lost a large share of their clicks, but they rarely produced demo requests anyway. Comparison pages have moderate AI exposure and stable clicks. Integration pages have almost no AI exposure, but their clicks fell sharply after a site migration.

    The report changes three decisions: stop defending glossary traffic as a KPI, keep comparison pages accurate because they are being cited and still convert, and treat the integration drop as a migration problem rather than an AI problem.

    The opt-out decision: this is where the data matters most

    Search Engine Land reported that as of 31 August 2026, Google rolled out a Search generative AI control to all websites worldwide. It lets a site block its content from AI Overviews, AI Mode and AI Overviews in Discover. Google’s statement, as quoted in that coverage, says that sites opting out “will not receive traffic or impressions from our generative AI features”, and that the control “will not be used as a ranking signal for search results outside of these generative AI Search features”. It applies sitewide, not per page.

    That makes the generative AI report the only first-party measure of what an opt-out removes. It still gives you only half the equation. You can see the impressions you would lose, but not the clicks, brand exposure or assisted conversions those impressions currently produce. When the control was first tested with UK publishers, Digiday’s reporting framed it as a choice publishers cannot safely use.

    A way to frame the decision with the data you have:

    • Cases for considering it: revenue depends on page views (ads, affiliate), most AI exposure sits on pages whose clicks are falling, and conversions on those pages have fallen with the clicks.
    • Cases against: AI exposure sits on pages with stable clicks, the business sells considered purchases where being named in an answer contributes to brand discovery, or the exposed pages are commercial pages that still convert.
    • Finer-grained controls: Google’s AI features documentation points to nosnippet, data-nosnippet, max-snippet and noindex to limit what is shown. These work per page or per section, but they also restrict regular search snippets, so they are blunt tools.

    If you do opt out, measure it properly. Because the control is sitewide, you cannot split-test it within one property. The best available design is a before and after comparison between high-AI-exposure page groups and low-AI-exposure page groups, which the opt-out should barely affect. Agree the metrics beforehand (organic clicks and conversions by group), run it for at least 8 to 12 weeks, and log core updates and site releases during the window. The published coverage does not say how quickly eligibility returns if you opt back in, so treat the decision as slow to reverse until Google documents it.

    Decisions the data can and cannot support

    Decision What the report adds What you still need What would change the call
    Where to invest content effort Which page groups carry AI exposure Clicks and conversions by the same groups Conversions stable despite falling clicks: lower priority than it looks
    Whether an organic decline is AI-related Whether declining pages have AI exposure at all Rank tracking and SERP feature history Declines concentrated in low-exposure pages: look elsewhere
    Opting out of AI features Size of AI impressions at stake, by page group and country Revenue model, conversion data, brand goals Exposure concentrated on pages that still convert: stay in
    Resetting stakeholder targets Evidence that visibility continues where clicks fall Agreement on what a non-click impression is worth Branded demand and direct traffic trending with AI exposure
    Market prioritisation Country split of AI exposure Knowledge of where each AI feature has launched Click declines limited to markets with heavy AI exposure

    Two supporting data sources help fill the gaps. The branded queries filter lets you check whether branded demand moves with AI exposure. That is weak evidence of an awareness effect, not proof. And Bing Webmaster Tools’ AI Performance report shows citations and a sample of grounding queries for Copilot and Bing’s AI summaries. It is a different engine, but it gives query-level hints that Google’s report withholds.

    What the report should not change

    Heavy AI exposure on a page is not a reason to rewrite it into short “extractable” chunks or add AI-specific markup. Google’s guide to generative AI features says none of that is needed and that optimising for these features is still SEO. The report tells you where you are visible. It says nothing about why, and it certainly does not confirm that a formatting change caused a citation.

    The practical position

    Stop treating the report as a standalone metric. Each month, join it to page-level clicks and conversions, group pages by template, and sort them into the four quadrants. Use the result to decide where content effort goes, to rule AI in or out when traffic falls, and, if you are a click-dependent publisher, to size the opt-out question with numbers rather than instinct. Plan a proper review once you have at least six months of data, because the first few weeks after a worldwide rollout are too short to separate a trend from a launch artefact.

    Need help with this on your own site?

    Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

    See SEO services

Add a reply

Replies are open to approved members. Log in or apply to join.