Google Search Console Generative AI Report & AI Controls (2026 Guide)
Updated 15 November 2026: the Search Console generative AI report still ships no click or query data, and on 2 November 2026 Google said it has nothing to announce on either. Earlier, October 20 2026: the report gained a traffic counterpart — a Penn/Northeastern randomised experiment with 1,100 US users found routing searches into AI Mode cut clicks to external sites by 18.8 percentage points, while hiding AI Overviews lifted external clicks by an estimated 8.8 points. Impressions and clicks now diverge by design, so this report is not a traffic proxy. New sections: the controlled evidence table, the AI-share metric (AI impressions ÷ web impressions per page per month), the 28-hour noindex case that cost ~55% of daily AI impressions, the 47.5%-from-one-URL property skew, and the fact that the report is still not exposed through the Search Console API — making a monthly manual export the only way to own your history Updated 15 November 2026: click data is still not coming. Search Engine Roundtable reported on 2 November that Gary Illyes said Google has nothing to announce on clicks or queries, while John Mueller called the pessimistic reading odd, so the honest planning assumption is an impressions-only report. The more useful November finding is Cyrus Shepard's one-site opt-out test (30 October): AI impressions were about 8% of total impressions, estimated CTR on them roughly 0.9%, and ordinary organic clicks also fell once AI features were disabled. Adds a question-by-question table of what the report can and cannot answer, next to the measured substitutes.
For the first time, GEO measurement moved out of third-party dashboards and into Google's own console. Through August 2026, Google rolled the Generative AI performance report and generative AI controls out broadly in Search Console — letting site owners both see how often their content appears inside AI Overviews and AI Mode, and decide whether Google may use it to ground AI answers at all. If you optimize for AI search, this is now the first surface you should check every week.
Updated September 2026: The report is no longer a rollout story — it is finished. Google completed worldwide availability for every verified property on 31 August 2026 (some publishers first saw it on 2 September), ending the UK-only test phase that began in June. Two operational changes matter more than the launch itself: as of 7 September 2026, Search Console combines AI Mode and AI Overview impressions rather than reporting AI Mode separately, and AI Mode rows can be filtered by user prompt — the first time Google has exposed the questions driving your AI exposure. One clarification, added 15 November 2026: AI Mode impressions are folded into the combined figure, but the prompt filter still applies to AI Mode rows — so the query-shaped insight survives while the surface-level split does not. Google still ships no click data, and John Mueller has been explicit about why. Earlier context still holds: Google reaffirmed AI Mode at 1 billion+ monthly active users on its Q2 2026 earnings call (July 22, 2026), and off-domain tracking remains mandatory because only 2.9% of brand citations point to the brand's own domain (Pratyush Kumar; 102,025 prompt responses across five engines).
Updated October 20, 2026: Two months of live data have changed how this report should be read. First, a randomised field experiment (University of Pennsylvania + Northeastern; 1,100 US Google users; reported by Search Engine Land on 23 September 2026) found that pushing users into AI Mode cut clicks to external websites by 18.8 percentage points — and hiding AI Overviews lifted external clicks by an estimated +8.8 points. Impressions and clicks now move in opposite directions, so this report's metric is not a traffic proxy at all. Second, the report is not exposed through the Search Console API, so every history you keep must come from a manual export. Section-level detail, a new AI-share metric, and a revised weekly playbook follow below.
Google's position, as reported by Search Engine Journal in September 2026, is that the generative AI performance report will keep evolving alongside Search, AI Mode and AI Overviews, that position counting cannot be an absolute truth, and that the goal is something useful to site owners rather than a fixed standard.
The shift is bigger than a new tab. For two years, GEO practitioners measured AI visibility with proxies — scraping ChatGPT answers, buying brand-mention tools, guessing from referral logs. Now Google exposes owned-domain AI impressions natively, and simultaneously hands publishers an opt-out that the UK's CMA forced into being. Below is what changed, what the numbers mean, and how to act on them.
What's new in Search Console (September 2026): Worldwide availability — finished 31 August 2026 for every verified property, ending the UK-only June test · Generative AI report — impressions across AI Overviews, AI Mode, and AI Overviews in Discover, by page / country / device / date; AI Mode and AI Overviews impressions are now combined (from 7 September 2026) · Prompt filtering — AI Mode rows can be filtered by the user prompts that triggered them · No clicks — still impressions-only, provisional data marked with a dotted line · Generative AI controls — property-wide opt-out at Settings → Crawling & Indexing → Search generative AI (not an organic ranking signal) · CMA deadline — page-level controls and click reporting due by March 2027 · Cloudflare AI-bot defaults from Sep 15, 2026 (search allowed; training & agent blocked on ad pages) · Pratyush Kumar 2026 study — only 2.9% of brand citations point to the brand's own domain.
The Generative AI performance report: what it actually shows
The report appears under the "Search results" menu in Search Console once Google has enabled it for your property (rollout accelerated through July–August 2026). It reports how your site is discovered and displayed inside Google's generative surfaces:
| Dimension | Reported? | What it tells you |
|---|---|---|
| Impressions (AI Overviews + AI Mode, combined) | Yes | How often your pages appear inside AI answers. Since 7 September 2026 the two surfaces are counted together — there is no AI Mode-only impression view. |
| Top pages | Yes | Which URLs earn the most AI visibility. |
| Country | Yes | Where your AI impressions concentrate geographically. |
| Device | Yes | Desktop vs. mobile AI exposure. |
| AI Overviews in Discover | Yes | A third surface in the same report — AI-generated summaries inside the Discover feed. |
| User prompts (AI Mode only) | Limited | AI Mode rows can be filtered by the prompts that triggered them — the closest thing to AI query data Google currently offers. |
| Clicks | No (yet) | Google does not yet report clicks, CTR, conversions or attribution for the generative AI report. |
Source: Google Search Central — Generative AI performance report documentation; ROI Revolution "August 2026 SEO & GEO News Recap."
The honest limitation: this report measures appearance, not attention. Because Google withholds click and query data, a page can show a large AI impression count while sending almost no referral traffic — the zero-click reality of AI answers. Use the report to confirm eligibility and spot which pages Google thinks are citable, then triangulate real impact with off-site tracking.
What changed on 7 September 2026: AI Mode impressions are folded in
The single biggest reporting change since launch is quiet and easy to miss. According to Google's Search Console documentation as of 7 September 2026, AI Mode impressions are no longer broken out separately — they are combined with AI Overview impressions in one generative AI figure. If you built a dashboard in August that split the two surfaces, that split stopped working.
- ▸ What you lose. Surface-level attribution. You can no longer say "AI Mode drove X% of our generative impressions" from first-party data alone; you need third-party prompt panels to separate them.
- ▸ What you gain. Prompt filtering. AI Mode rows accept a user-prompt filter, letting you inspect which questions triggered your inclusion — currently the only query-shaped signal Google exposes for generative surfaces.
- ▸ Why it is structurally sensible. AI Mode and AI Overviews draw on the same Google Search index rather than a separate AI index, so a single combined counter matches how Google says the systems work.
- ▸ What comes next. The UK Competition and Markets Authority has set Google a March 2027 deadline to ship page-level AI controls and proper click reporting. Today's property-wide toggle is a first iteration, not the end state.
Practical handling: annotate any generative-AI reporting you hand to stakeholders with "AI Overviews + AI Mode combined" from 7 September onward, and keep a manual AI Mode prompt sample so you retain some surface-level context the console no longer provides.
Updated 15 November 2026 — clicks are not coming, and a one-site experiment shows why you should stop waiting for them. On 2 November 2026, Search Engine Roundtable reported that Google's Gary Illyes answered a claim that click and query data were headed for the report by saying the team is interested in giving more data, data that makes sense, but has nothing to announce yet — while John Mueller called the pessimistic reading of that exchange odd. Read it as a non-commitment in either direction, and plan on the report staying an impressions ledger. The more useful November number came from outside Google: Cyrus Shepard used the generative AI opt-out control to remove one site from AI Overviews and AI Mode and compared the periods in and out. AI impressions were about 8% of the site's total impressions, estimated CTR on those impressions was roughly 0.9%, and — the surprising part — regular organic clicks also fell once AI features were switched off. It is one site with a subtracted CTR, so treat it as a hypothesis worth testing, not a benchmark; but it kills the folk plan of opting out to protect blue-link traffic.
November 2026: what the impressions-only design actually costs you
Three numbers published in the last six weeks now fit together, and they describe the measurement problem this report leaves unsolved rather than a Google defect.
| Question | Can the generative AI report answer it? | Measured substitute (with source) |
|---|---|---|
| Did my page appear inside AI Overviews or AI Mode? | Yes — impressions by page | Nothing needed. This is the report's one native metric. |
| How many clicks did AI features send? | No — clicks sit inside Web-search totals | Estimate by subtraction, as Shepard did: ~0.9% CTR on AI impressions, AI impressions ≈ 8% of total impressions (one-site case study, October 2026) |
| Which queries triggered that appearance? | No — no query dimension | AI Mode prompt filtering (partial) plus a hand-maintained tracked-prompt panel |
| Am I cited more or less than competitors? | No — no cross-site view | A public citation index, read as a dated fraction (numerator over denominator, with the release id) rather than as a rank |
| Would opting out protect my organic clicks? | Not answerable from the report | Shepard's test says no — organic clicks fell with AI features disabled on that site |
Sources: Search Engine Roundtable (2 November 2026, Illyes on click data; 30 October 2026, Shepard opt-out test); Google Search Console Help, Generative AI performance report.
The pattern is stable enough to design around: the console tells you whether you were shown, and everything about whether that showing was worth anything has to be measured somewhere else. That is not a reason to ignore the report — impressions by page remain the cheapest way to find out which of your URLs Google considers citable, and a page with growing AI impressions is the best candidate for the next GEO rewrite. It is a reason to stop treating the report as the scoreboard.
One page in one account took half the AI impressions (1,190 of 2,330), and the country splits did not follow organic search: 908 impressions from Nigeria against 552 from the United Kingdom, on a site whose audience is Nigerian businesses. Export the report monthly and keep your own history — the API does not expose this view, so anything you do not export is lost.
The generative AI controls: opt in, opt out, or inherit
Running in parallel with the report, Google's generative AI controls let you decide whether your site's links and content appear in AI Overviews, AI Mode, and AI Overviews in Discover. Find them at Settings → Crawling & Indexing → Search generative AI. One constraint to know before you plan around them: the control is property-wide, not per page — you cannot exclude one section while leaving the rest eligible. The three options:
- ▸ Include — Google may use your site to ground AI responses and show links to your content.
- ▸ Exclude — Google will not use your pages to ground AI answers or link to them.
- ▸ Inherit — child properties follow a parent property's setting (overridable).
Critical nuance: Google states this setting is not used as a ranking signal for ordinary organic search. Excluding your site from generative AI does not change your "10 blue links" rankings — but it does remove you from AI answers entirely. For most GEO strategies, allowing Google to cite your site is the right call: being cited at the top of an AI Overview raises the chance of a click and helps you own your narrative. Reserve the opt-out for cases where appearing in AI answers carries legal, regulatory, or brand-safety risk.
The control's origin is regulatory, not charitable. On 3 June 2026, the UK Competition and Markets Authority imposed a conduct requirement giving publishers "effective tools to prevent their content being used to power AI features in search, such as AI Overviews," obliging Google to attribute publisher content with clear links inside AI-generated results and to let publishers opt out of model fine-tuning separately. Google has nine months to implement, putting the deadline at March 2027 — and the outstanding items are exactly the two gaps publishers complain about most: page-level (rather than property-wide) AI controls and click reporting. The Search Console control available today is the user-facing expression of that requirement, rolling out UK-first and completing worldwide availability on 31 August 2026.
October 2026: the first controlled evidence that impressions and clicks diverge
Through September, the objection to the generative AI report was theoretical: it counts impressions but not clicks. A preregistered randomised field experiment has now turned that objection into a measurement problem. Researchers at the University of Pennsylvania and Northeastern University ("AI in Search Reduces Publisher Referrals Without Improving User Experience," released August 2026; covered by Search Engine Land on 23 September 2026) used a browser extension to randomly assign 1,100 US Google users to one of three conditions for one week: no AI at all, Google as normal, or every search routed into AI Mode.
| Measure (AI Mode vs. Google as normal) | Change | What it implies for the GSC report |
|---|---|---|
| Clicks to external websites | −18.8 pp | Rising AI impressions can coexist with falling referrals — by design, not by error. |
| Clicks to Reddit / news sites / Wikipedia | −21.2 / −12.5 / −9.9 pp | Information-shaped destinations lose most; transactional pages lose least. |
| External clicks when AI Overviews are hidden | +8.8 pp (est.) | The AI layer removes clicks without adding satisfaction — a structural, not temporary, effect. |
| Use of rival search engines | +11.2 pp | Defection rises when AI Mode is forced: Google's own AI features carry retention risk. |
| Google search sessions per day | −0.92 | Fewer sessions means fewer impression opportunities downstream as well. |
Source: University of Pennsylvania + Northeastern University, "AI in Search Reduces Publisher Referrals Without Improving User Experience" (preprint, August 2026); reported by Search Engine Land, 23 September 2026. Caveats from the authors: AI Mode represented only 0.6% of baseline searches, the sample skewed younger and more educated, and Google changed its page code mid-study so only about half of AI Overviews were successfully hidden — meaning the +8.8 pp uplift is likely an underestimate.
"If traffic falls while rankings and impressions hold steady, it is easy to assume something has gone wrong with the SEO. In reality, the search results page is simply answering more of the question itself."
The operational consequence is blunt: stop treating rising generative AI impressions as evidence that visibility is converting into demand. Report them next to clicks and sessions, never instead of them. Where impressions hold and clicks fall, the result page changed — not your rankings.
The AI-share metric: the one number the report lets you build
Google confirms that ordinary web impressions already include appearances in AI Overviews and AI Mode — the generative AI view isolates a subset, it is not an additional channel. That detail is also the report's most useful property: because the AI numbers sit inside the web numbers, you can divide one by the other for each page.
AI share = generative AI impressions ÷ web search impressions (same page, same month)
- A page whose AI share climbs while clicks stay flat is being shown inside answers more often without gaining visits — the exact failure mode the Penn experiment predicts.
- A 28-hour site-wide noindex at one publisher cut daily AI impressions by roughly 55% at the low point (Value Add VC, September 2026) — indexing is a prerequisite, and AI impressions are a fast index-health alarm.
- One audited site saw AI impressions rise from 411 (May 2026) to 5,650 (August 2026), roughly 14×, while publishing almost no new content — existing pages simply got picked up more often. Growth in this report often reflects retrieval changes, not content changes.
- A single URL can carry nearly half the total: at taismo.de, one glossary page produced 47.5% of all AI impressions — and it faded without ever ranking badly. Always split the property total by page before drawing conclusions.
- Content type moves the ratio: a comparison article earned 39 AI impressions per 100 web impressions versus 5 for glossary pages in the same property — comparison content is pulled into answers far more often than definitions.
One more mechanical constraint to plan around: as of October 2026 the Search Console API still exposes only web, image, video, news and Discover search types — there is no AI Overviews or AI Mode type. No tool can pull this report automatically, so a monthly manual export is the only way to own your history. Export the Search view (not Discover) and store the file; the report's own retention is not your baseline.
Why impressions are not the whole story: the 2.9% finding
Even with Google's report live, it only counts your own domain. And the own-domain slice of AI visibility is surprisingly small. A 2026 large-scale baseline (Pratyush Kumar; capston.ai analysis) examined 102,025 prompt responses and 149,912 citations across 102 brands on ChatGPT, Gemini, Perplexity, Claude, and Grok (March–May 2026). Two findings reframe the work:
| Finding | Value |
|---|---|
| Citations pointing to the brand's own domain | 2.9% |
| Citations to corporate + third-party brand pages | ~75% |
| Ranked "best-of" listicles as share of all citations | 21.0% |
| Brand-stature ladder (household / mid-market / niche first-run appearance) | 72.9% / 43.6% / 11.4% |
The consequence is uncomfortable for anyone selling on-site optimization alone: engines mostly cite pages about you that you do not own. Search Console shows your 2.9%; the other ~97% lives on review sites, roundups, and third-party coverage. A complete GEO measurement program pairs the native report with off-domain brand-mention tracking (see our guide on tracking your brand in AI answers). One more wrinkle from the same study: sentiment flipped in 45.5% of runs versus 6.8% for mere mention status — how you are framed is roughly 6.7× noisier than whether you appear at all.
Cloudflare's September 2026 AI-bot defaults
Crawler access is the gatekeeper beneath all of this. From September 15, 2026, Cloudflare applies new default posture for AI bot traffic, split into three categories:
- ▸ Search bots — allowed by default. These are the crawlers that help your content appear in AI answers (e.g., Google-Extended, OAI-SearchBot). Keep them allowed.
- ▸ Training bots — blocked by default on pages that display ads.
- ▸ Agent bots — blocked by default on pages that display ads.
The GEO takeaway: the bots that drive AI citations stay on by default, so most sites need no change — but if you run Cloudflare and have been blocking bots wholesale, now is the time to confirm search crawlers are explicitly allowed. In response to the announcement, Google's John Mueller reiterated that Google does not use llms.txt or llms-author.txt files — another signal that crawler access and structured content, not self-described summaries, govern AI eligibility.
How to read your AI report: a 5-step weekly playbook
- 1.Confirm the report is enabled
Open Search Console → "Search results" → look for the generative AI dropdown. If absent, it is coming; check again weekly. Verify the AI opt-out control is set to "Include" unless you have a brand-safety reason not to.
- 2.Rank pages by AI impressions
Your top AI-visible URLs are the pages Google's models already trust. Protect them: keep dateModified fresh, verify Schema.org, and ensure Google-Extended can crawl them.
- 3.Cross-check against referral traffic
High AI impressions with low referrals is normal under zero-click. Do not panic — but do treat the impression count as a visibility score, not a traffic number.
- 4.Audit off-domain citations
Since ~97% of your brand's AI citations sit off your domain, track review sites, roundups, and "best-of" listicles that mention you (ranked listicles carry 21% of all citations). Influence them with accurate, citable data.
- 5.Watch sentiment, not just mentions
Framing flips in 45.5% of runs. One negative-looking AI answer is noise, not a crisis — but a pattern of mis-framing is a real GEO problem worth correcting at the source.
Common mistakes with the new controls
- ▸ Treating AI impressions as clicks — The report has no click or query data yet; pair it with referral logs and brand-mention tools.
- ▸ Opting out by default — Excluding your site removes you from AI answers and is not an organic ranking signal; only opt out for genuine risk reasons.
- ▸ Ignoring off-domain footprint — 97% of brand citations live elsewhere; on-site optimization alone misses most of your AI visibility.
- ▸ Blocking search bots at the edge — A Cloudflare or security rule that blocks AI search crawlers silently opts you out of every AI citation, regardless of Search Console settings.
Frequently asked questions
What does the Google Search Console Generative AI report show?
Worldwide availability finished on 31 August 2026. The Generative AI performance report in Search Console shows your site's impressions across Google AI Overviews, AI Mode, and AI Overviews in Discover, broken down by page, country, and device. Since 7 September 2026 AI Mode and AI Overview impressions are combined rather than reported separately, and AI Mode rows can be filtered by user prompt. The report still contains no click data, so it measures AI visibility (appearance), not traffic.
Can I stop Google from using my content in AI Overviews and AI Mode?
Yes. Google's generative AI controls (rolled out UK-first from mid-2026, expanding globally) let you choose whether your site's links and content appear in AI Overviews, AI Mode, and AI Overviews in Discover. Opting out prevents Google from using your pages to ground AI answers. Google states this setting is not used as a ranking signal for ordinary organic ("10 blue links") search.
Do Search Console AI impressions equal clicks?
No. AI impressions measure how often your content appears inside an AI answer; they do not count clicks, because Google does not yet report click or query data for the generative AI report. This matters because AI Overviews and AI Mode are heavily zero-click: a high impression count can coexist with little or no referral traffic, so pair the report with off-site brand-mention tracking.
How much of my AI citations come from my own website?
Very little. A 2026 large-scale study (Pratyush Kumar; 102,025 prompt responses and 149,912 citations across 102 brands on ChatGPT, Gemini, Perplexity, Claude, and Grok) found only 2.9% of all citations point to the brand's own domain — about 75% go to corporate and third-party pages, and ranked "best-of" listicles alone carry 21% of citations. Optimizing only your own site misses the majority of your AI footprint.
What changed with Cloudflare AI bot defaults in September 2026?
From September 15, 2026, Cloudflare set new default posture for AI bot traffic: search bots (the kind that help you appear in AI answers) are allowed by default; training bots are blocked by default on pages that display ads; and agent bots are blocked by default on ad-displaying pages. Site owners can override these per category. The takeaway for GEO: keep search crawlers allowed so your content stays eligible for AI citation.
Can I see how much of my generative visibility came from AI Mode specifically?
Not natively. Since 7 September 2026, Search Console combines AI Mode and AI Overview impressions instead of reporting AI Mode separately, so there is no AI Mode-only impression, click, CTR, conversion or attribution view. Workarounds: filter AI Mode rows by user prompt to see which questions triggered inclusion, or use a third-party prompt panel to measure the two surfaces separately.
Is Google adding click data to the generative AI report?
Not as of 15 November 2026, and Google has committed to nothing either way. On 2 November 2026, Search Engine Roundtable reported Gary Illyes saying the team is interested in giving more data that makes sense but has nothing to announce yet, while John Mueller called the pessimistic reading of that comment odd. Plan as if the report stays impressions-only: export it monthly, keep your own history, and estimate AI clicks by subtraction. In the one published opt-out test (30 October 2026), AI impressions were about 8% of total impressions with an estimated CTR of roughly 0.9% on them.
Does opting out of AI Overviews protect my organic clicks?
On the only site where this was tested publicly, no. Cyrus Shepard used the generative AI opt-out control to remove one site from AI Overviews and AI Mode and compared the periods in and out: AI impressions were about 8% of total impressions, estimated CTR on those impressions was roughly 0.9%, and ordinary organic clicks also fell while AI features were disabled — what he called a SERP halo effect. One site plus a subtracted CTR means this is a hypothesis, not a benchmark; it does mean "opt out to protect clicks" is not a safe default.
Can rising AI impressions still come with falling traffic?
Yes, and it is now causal rather than correlational. A preregistered randomised experiment from the University of Pennsylvania and Northeastern University (1,100 US Google users, reported 23 September 2026) found that routing searches into AI Mode cut clicks to external websites by 18.8 percentage points, while hiding AI Overviews raised external clicks by an estimated 8.8 points. The recommendation: report generative AI impressions beside clicks and sessions, never instead of them, and treat rising impressions plus flat clicks as a change in the result page rather than a ranking loss.
Can I pull the generative AI report data through the Search Console API?
No. As of October 2026 the Search Console API still only offers web, image, video, news, Google News and Discover as search types, with no AI Overviews or AI Mode type. Because no third-party tool can retrieve it automatically, the only way to keep a history is a manual monthly export. Export the Search view (not Discover) and store the file, then compute each page's AI share as generative AI impressions divided by web search impressions for the same page and month.
Is the Google AI opt-out control page-level or site-wide?
Site-wide, for now. The control sits at Settings → Crawling & Indexing → Search generative AI and applies to the whole verified property, so you cannot exclude one section while leaving the rest eligible for AI answers. The UK Competition and Markets Authority has given Google a March 2027 deadline to introduce page-level AI controls alongside click reporting for these surfaces.
Related GEO guides
References: Google Search Central — Generative AI performance report and generative AI controls documentation (2026). · ROI Revolution — "August 2026 SEO & GEO News Recap" (new Search Console reports, generative AI controls, Cloudflare AI-bot defaults). · capston.ai — "AI Search Watch" (week of Aug 10, 2026): Pratyush Kumar brand-visibility baseline, arXiv:2606.20065, 102,025 prompt responses / 149,912 citations / 102 brands / 5 engines (Mar–May 2026). · UK Competition and Markets Authority — conduct requirement on publisher AI-content controls (3 June 2026). · Search Engine Land — Pieter Serraris, ChatGPT language-bias study (ChatGPT-User cites English pages 2.6× more than Googlebot, 2026). · Google Search Central — John Mueller on llms.txt (2026). · Google — Q2 2026 earnings call (July 22, 2026): AI Mode reaffirmed at 1 billion+ MAU, expanding the Generative AI report's measurable surface. · Searcle, "How Google's Conversational Search Works" (status note dated 7 September 2026) — Search Console combines AI Mode and AI Overview impressions rather than reporting AI Mode separately; no AI Mode-only click, CTR, conversion or attribution view exists. · Search Engine Land — coverage of the 31 August 2026 worldwide rollout and its link to the CMA's March 2027 deadline for page-level AI controls and click reporting. · John Mueller, Search Advocate at Google, public comments on generative AI performance reporting (September 2026) — reporting aims to be useful rather than a "written-in-stone absolute truth for position counting." · Intelligency Group / Sited — summaries of the Penn + Northeastern randomised field experiment (1,100 US users; AI Mode −18.8 pp external clicks; no-AI group +8.8 pp; rival engines +11.2 pp; −0.92 sessions/day), reported by Search Engine Land 23 September 2026. · taismo.de — "Search Console AI Report" (30 June – 29 September 2026): one glossary URL produced 47.5% of AI impressions; baseline doubled from 50–280 to 300–600/day; comparison pages 39 AI impressions per 100 web impressions vs 5 for glossary pages. · Value Add VC — AI search traffic data (October 2026): 1.91M AI impressions vs 537 AI referrals; 28-hour site-wide noindex cut daily AI impressions ~55%; AI impressions 411 → 5,650 (May → August 2026). · Citable — the Search Console API still exposes only web, image, video, news, Google News and Discover; generative AI report available as manual export only (October 2026).
Want to check your site's GEO readiness?
Run the 27-point GEO auditRelated articles
How to Measure GEO Visibility: 4 Metrics + the Retrieval Gate (2026)
Updated 18 November 2026 with the frame in which this page lost its own query. For its head query on 13 November 2026, two full-text restatements of this page (hosted on ima.qq.com) occupied positions 1 and 2 while geoaura.world did not appear in the result set at all - the URL surfaced only as link text inside the restatement. That separates three things usually bundled together: authorship, attribution and occupancy, which means a mention rate and a presence rate must be counted separately (1.0 mention, 0.0 presence on the same query, same day). The page now opens with a dated original-source statement and a four-step response that does not include chasing the restatement: date the original in its first screen, keep the canonical attribution sentence in body text rather than metadata, keep superseded versions in one log, and re-measure the same query for three days before concluding anything. Also shows why its own external reading is now interpolated rather than written: the 17 November 2026 frame reads 15 citations in 40,578 observed answers (0.0370%), last cited 12 November — a count that has not moved while the corpus grew 767 (1.93%) — so any page that printed a single hard-coded denominator became wrong within 48 hours. Retains the retrieval-gate methodology (PressGEO 33.1% brand-named versus 0.0% on 643 non-brand probes), the crawl-versus-citation telemetry (856 crawls producing zero citation referrals against 868 producing 239), the four metrics, and the weekly cadence.
8 Best AI Search Visibility Tools for GEO Tracking (2026)
AI search visibility tools split into two products that share one query: software that measures your visibility, and agencies that sell you the visibility itself. Updated 1 November 2026 with a six-row table that tells the two apart, plus the numbers that settle the choice: DerivateX found 12.2% of 1,259 Google AI Overview citations on B2B software queries pointed at the recommended product's own site, with 30 of 100 queries recommending products while citing none of their pages; Ranqo Labs classified 149,912 citations and found 2.9% pointing at a brand's own domain. Also carries the eight-tool comparison (Omnia, Semrush, Profound, Peec AI, Backlinko, AthenaHQ, Otterly.AI, AI-Search), cost per 1,000 tracked responses, the four questions that separate a measurement provider from a content provider, and the red-flag list for 2026.
How to Track Your Brand in AI Answers (Step-by-Step)
Updated 17 November 2026: same-day repeat-run studies split the two metrics a brand panel usually blends. Across three runs of the same question, only 15% of the sites ChatGPT cited appeared every time, against 55% for Google AI Overviews and 96% for Perplexity, while 46% of the brands it named did - so mention is the series to run trends on and citation is the one that needs more runs per engine. Cited sites return 48% an hour later and 37% after two weeks on ChatGPT, against 99% and 62% on Perplexity. New: the three-field model (mention, citation, presence), the reason 46% of ChatGPT brand mentions carry no cited page, and a first-party case of 1.0 mention with 0.0 presence on the same query. Previously October 20 2026: two October panels settle whether one engine is enough. Pineprompt tracked 18,792 answers across seven engines for 14 days and found only 12.9% of named brands appeared on every engine while 46.2% appeared on a single engine only, with day-to-day churn up to ~69% on Perplexity; LLM Scout found ChatGPT, Claude, Gemini and Perplexity disagreed on whether to name a tracked brand in 27.2% of 749 identical buying questions. New: an engine-sampling table, a disagreement-as-noise-floor rule, and the 5-step methodology (query set, multi-run, extraction, aggregation, trend) across ChatGPT Search, Perplexity, Google AIO, Claude and Gemini.