--- title: "You're Measuring Google AI Overview Performance Wrong | Chad AI" description: "Google counts AIO impressions only when links scroll into view — not on page load. If you're reading Web reports, not Generative AI reports, your AIO data is wrong. Here's how to fix your measurement before it corrupts your GEO strategy." canonical: https://trychad.ai/blog/google-aio-impression-counting-scroll-to-view/ --- BENCHMARKGEO · Aug 11, 2026 · 6 min read · By Alexander Shoucair # You're Measuring Google AI Overview Performance Wrong // TL;DR Google Search Console counts AI Overview impressions only after a link scrolls into view — not on page load. Your Web report mixes AIO data with organic in ways that make both look wrong. Google launched a dedicated Generative AI performance report in June 2026 that separates the two cleanly. If you haven't switched, every AIO metric you're reporting is unreliable. Most teams tracking Google AI Overview performance are reading the wrong report. The measurement rule changed, the right tool launched, and nobody noticed — so the data everyone is acting on is structurally broken. Here is the exact rule, confirmed by Google and reported by Search Engine Journal on August 4, 2026: an AI Overview impression is counted only after a link within the AIO has been scrolled into view or expanded. Page load does not trigger the count. A user who lands on a search result page and scrolls past the AIO without stopping — no impression is counted for any URL inside it. That one rule changes your numbers substantially. If your AIO content sits below the fold, and most users don't scroll into it, your impression count understates actual exposure. If you are running campaigns and tracking AIO presence as a success metric, you may be crediting pages that are technically appearing but never being seen. ## What the counting rules actually say Google's AIO impression counting works at two levels, and the distinction matters for multi-URL sites: **Property level:** If two URLs from the same site appear in a single AI Overview, the property receives one impression — not two. The AIO is treated as a single unit. You cannot double-count by having two pages cited. **Page level:** Each URL within an AIO may receive its own impression. So if page A and page B both appear in the same AIO, page A gets one impression and page B gets one impression — but the property-level count still shows one. **Position:** All links within a single AIO share a single position. An AIO that contains five cited URLs does not give each URL a separate position. They all hold position 1 (or wherever the AIO ranks on that SERP). This means property-level AIO impression data in your Web report is already a significant undercount relative to how many pages you have cited — and the position field tells you nothing about your citation frequency within the AIO itself. ## Why the Web report is the wrong place to look The Search Console Web report combines AI Overview appearances with traditional organic results in ways you cannot separate by filtering. A URL that ranks #3 organically and also appears in an AIO for the same query generates both organic and AIO signals in a single row. Impressions, clicks, and CTR in that row are a blend. There is no filter in the Web report that isolates AIO traffic. You cannot segment it out. You cannot see whether a click came from the AIO or from the organic blue link below it. This is not a new limitation. It is the reason Google launched a separate report entirely. ## The Generative AI performance report — what it gives you Google launched dedicated Generative AI performance reports in Search Console on June 3, 2026. These reports cover both AI Overviews and AI Mode, and they separate AI-generated surface traffic from traditional organic results. What you can see that you cannot see in the Web report: - Impressions generated specifically from AI Overview appearances, counted under the scroll-to-view rule - Clicks from AIO surfaces, distinct from organic clicks for the same URL - CTR for AI surfaces separately — so you can see whether AIO appearances are generating click-through or operating as zero-click surfaces - AI Mode performance separately from AIO performance (Google treats these as distinct citation surfaces with different behaviours) - Query-level data: which search terms triggered an AIO citation for your pages The report is in Search Console under Performance → Search type → AI Overviews. If your account has sufficient data volume, it will also appear as a dedicated card in the Performance overview. ## Why this matters for GEO measurement Generalised Engine Optimisation strategy depends on knowing whether your content is being cited. If your measurement is broken, your strategy is flying blind. Three specific errors follow from using the Web report for AIO measurement: **1. You misattribute AIO improvements to organic SEO.** If a content update drives an AIO citation, the Web report registers increased impressions on that URL — but you cannot tell whether the improvement came from the organic ranking or the new AIO citation. You credit the wrong change. **2. You cannot diagnose zero-click behaviour accurately.** AIO surfaces regularly generate impressions without clicks — users get the answer from the AI summary and do not visit your site. In the Web report, this appears as an unexplained CTR drop for high-impression pages. In the Generative AI report, it is visible as a distinct AIO CTR that you can benchmark against organic CTR separately. **3. You cannot validate whether GEO work is generating AIO citations.** If you optimise content for citability and want to measure whether it is getting cited in AIOs, the only reliable measurement surface is the Generative AI performance report. The Web report cannot confirm a citation appeared. ## What to do Switch your AIO measurement to the Generative AI performance report. This is not optional if you want accurate data. Step 1 — Access the report In Google Search Console: Performance → Search type dropdown → select “AI Overviews”. If you have sufficient query volume, this will be populated. Smaller sites may see limited data in the first 60–90 days. Step 2 — Establish a baseline from today Do not try to retroactively reconcile Web report data with Generative AI report data. The counting methodology is different. Start fresh: record your current AIO impression count, AIO click count, and AIO CTR as of today. That is your baseline. Compare future periods to this, not to pre-June historical data. Step 3 — Track the right metrics AIO impressions (scroll-activated) tell you visibility. AIO clicks tell you whether the citation generates click-through. AIO CTR is the ratio — for most informational queries this will be low (under 3%), which is normal for zero-click surfaces. A high AIO impression count with near-zero CTR is not a problem; it means you are providing the answer directly, which is how AI Overviews are supposed to work. Step 4 — Use query-level data to refine content strategy The Generative AI report shows which queries are generating AIO citations for your URLs. This is the most valuable data in the report. Pages that receive AIO citations on high-volume queries are your benchmark pages — use them as templates for content you want to optimise for citability. Pages that rank organically but do not appear in AIOs for the same queries are your optimisation targets. Step 5 — Stop reporting AIO performance from the Web report Update any dashboards, client reports, or internal tracking that is pulling AIO data from the Web report. Flag any historical data from before June 2026 as “pre-dedicated-report” — it was measured differently and is not directly comparable. ## Chad's take The uncomfortable truth here is that most GEO reporting being presented to clients right now is inaccurate. Not because anyone is being dishonest — because the tool everyone defaulted to was the wrong tool, and the right tool launched quietly in June with no fanfare. The agencies and teams who are tracking AIO performance in the Web report are showing clients numbers that blend organic and AI traffic in ways that cannot be separated. They are crediting wins to the wrong changes and missing losses they cannot see. The fix takes five minutes: switch to the Generative AI performance report, establish a clean baseline, and stop reporting from the old surface. If you are in a client-facing role and you are not doing this, you are building your GEO strategy on broken data. ## Sources - **TryChad.ai** — Our own `dateModified` rule: derived automatically from commit history it made all 20 posts falsely claim same-day revision, and was reverted within the hour, August 2026. Primary source. The same measuring-the-wrong-thing failure this article documents in Search Console. - **Search Engine Journal (Carolyn Shelby)** — Google AI Overview impression counting rules: scroll-to-view activation, property vs. page level impression counting, August 2026. [searchenginejournal.com](https://www.searchenginejournal.com/google-ai-overviews-impressions/) - **Google Developers** — Generative AI performance reports in Search Console (AI Overviews and AI Mode), launched June 3, 2026. [developers.google.com](https://developers.google.com/search/docs/appearance/ai-overviews/performance-reports) - **Keywords Everywhere** — AI Overviews tracker and impression counting confirmation, August 2026. [keywordseverywhere.com](https://keywordseverywhere.com/news/ai-overviews) // want clean AIO measurement for your site? Chad runs GEO audits that include AIO citability scoring, Generative AI report setup guidance, and a full measurement framework. Book a walkthrough. [Book a walkthrough →](https://trychad.ai/book)