LaCleoThe short answer: three things changed in 2026. Google made AI visibility measurable by launching Search Generative AI performance reports in Search Console on 3 June. It published its first official guidance on generative AI features on 15 May, stating that optimising for AI search is still SEO. And the commercial layer arrived, with advertising inside ChatGPT from 9 February and agent checkout protocols going live across major retail platforms. The discipline has stopped being speculative and started being reportable.
The numbers behind it, at a glance:
| Signal | Where it stands |
|---|---|
| Google queries triggering an AI Overview | 48% of tracked queries, up from 6.5% a year earlier |
| US Google searches ending without a click | 68%, January to April 2026 |
| Conversion advantage of AI-assistant referrals | +42% versus non-AI traffic, March 2026 |
| Correlation between a top-ten ranking and an AI Overview citation | 38%, down from 76% |
Sources: BrightEdge, SparkToro, Adobe Digital Insights, Ahrefs. Full list at the foot of this article.
1. Google finally gave you AI visibility data
For two years the honest answer to "are we appearing in AI Overviews?" was a shrug and a third-party sample. That ended on 3 June 2026, when Google launched Search Generative AI performance reports inside Search Console. There are two views: one covering generative AI features in Search, meaning AI Overviews and AI Mode, and one covering generative AI features in Discover.
Read the small print before you build a report around it. The data shows impressions only, split by page, country, device and date. There are no clicks, no click-through rate, no query-level data and no API access. History begins on 18 May 2026, so year-on-year comparison is not possible yet. Rollout has been phased, starting with a subset of properties, with UK site owners in the first wave. And critically, this is a breakout rather than new data: those impressions were always inside your aggregate Performance totals, so nothing about your headline numbers has changed.
Alongside the reports Google shipped a control that lets a site block its content from appearing in AI Overviews, AI Mode and Discover AI features, honoured from 17 June 2026. Google has been explicit that opting out means forgoing impressions and traffic from those surfaces, and that the setting does not cover the Gemini app. For almost every commercial site this is a toggle to leave alone.
Search Console now tells you whether you appeared. It still will not tell you what it was worth.
2. Google published its own guidance, and it says GEO is SEO
On 15 May 2026 Google Search Central published its first official document on optimising for generative AI features, filed under a new "Generative AI fundamentals" section. Its central claim is that AI Overviews and AI Mode are built on the same core ranking and quality systems as the rest of Search, so "optimizing for generative AI search is optimizing for the search experience" and remains SEO by another name.
The guide includes a mythbusting section that names several fashionable tactics as unnecessary: you do not need new machine-readable files, AI-specific text files, or special structured data types to appear in AI features. Google also extended its spam policies to cover generative AI responses in Search, and called out mass-producing a page for every query variation as scaled content abuse.
Two working conclusions follow. First, the technical foundations still decide everything: crawlability for the right user agents, clean canonicals, fast rendering, sound internal linking, verified Search Console properties. Second, the differentiator moves upstream to whether a page contains something a model cannot synthesise from ten other pages. Practitioners running both disciplines together report that roughly four fifths of what helps SEO also helps GEO, and the remaining fifth is where the specialist work sits.
3. The click economy kept shrinking
Zero-click behaviour is no longer a talking point, it is the base case. SparkToro measured 68% of US Google searches ending without a click across the first four months of 2026, up from 60% in 2024. AI Overviews now trigger on roughly 48% of tracked queries, against 6.49% a year earlier. On business-to-business technology queries the trigger rate has been measured at around 82%.
The results page absorbed the answer:
| Measure | Earlier period | Now |
|---|---|---|
| AI Overview trigger rate (BrightEdge) | 6.5% (early 2025) | 48% (2026) |
| Zero-click searches, US (SparkToro) | 60% (2024) | 68% (Jan–Apr 2026) |
The practical consequence is that impressions and clicks have decoupled as measures of the same thing. A page can be doing more commercial work than ever while its click curve flattens, because the brand is being read inside the answer rather than visited. Reporting that still treats sessions as the only proof of visibility will systematically undercount the channel.
4. AI traffic is small, and it is the best traffic you have
Across roughly 13,770 domains studied by Conductor, visible AI referrals average about 1.08% of sessions. That number is why most teams ignore the channel, and it is the wrong number to look at.
Adobe Digital Insights found that in March 2026, visitors arriving from AI assistants converted 42% better than non-AI traffic. Twelve months earlier the same channel converted 38% worse, which is close to an eighty-point swing in a year. The pattern repeats across independent datasets at different magnitudes. Ahrefs found AI search accounted for 0.5% of its traffic but 12.1% of signups. Semrush measured a 4.4x conversion multiple across industries. A study of 94 ecommerce sites recorded a narrower but real 1.81% against 1.39% for non-brand organic.
Measured conversion advantage of AI referral traffic, 2025 to 2026:
| Study | Multiple vs baseline |
|---|---|
| Ahrefs, own traffic | 23.0x |
| Microsoft Clarity, publishers | 17.0x |
| Semrush, cross-industry | 4.4x |
| Adobe, retail Q1 2026 | 1.42x |
| Visibility Labs, 94 ecommerce sites | 1.31x |
The spread is wide because each study defines its baseline differently. The direction is consistent across every published dataset: AI-referred visitors arrive later in the decision, having already seen the brand compared against alternatives.
The engine mix is also broadening. ChatGPT's share of worldwide AI referral traffic fell from about 79.7% in 2025 to 76.0% in 2026, even as its absolute referral share to websites rose 36.7% month on month in May alone. Gemini, Perplexity and Copilot are taking measurable slices. Tracking one engine is now a sampling error rather than a strategy.
5. Citations concentrate, and page-one ranking no longer guarantees them
Two findings should reshape content planning. First, citation supply is extremely concentrated: in an analysis of 1,000 AI Overviews, the top 1% of cited domains, roughly a dozen sites, captured 47% of all citations, with the next 9% taking a further 31%. Second, the old proxy has weakened. Across 863,000 keywords, the correlation between holding a top-ten ranking and being cited in an AI Overview fell from about 76% to 38%, because Google fans a query out into related sub-queries before retrieval.
Where AI answers pull their sources (share of AI answers referencing each domain, across 150,000 citations and 5,000 keywords):
| Domain | Share of AI answers |
|---|---|
| 40.1% | |
| Wikipedia | 26.3% |
| YouTube | 23.5% |
| LinkedIn (B2B skew) | 13.5% |
| Brand-owned domains | ~9% (indicative, varies sharply by category) |
User-generated and reference domains dominate the citation pool. Sources: Semrush 150,000-citation analysis; Google AI Mode LinkedIn share per 2026 citation index research.
Page-level signals that correlate with being cited are unglamorous and consistent. Domain authority is the single strongest correlate. Pages over 2,500 words are cited around 1.6 times more often than pages under 800. Pages containing at least one named, attributed source are cited roughly 2.1 times more often than pages with none. Across a 25,337-citation study, comparison pages earned the highest citation rate of any page type at 1.87 citations per retrieval, ahead of articles, listicles and product pages.
Overlap between engines is low. Only about 11% of domains are cited by both ChatGPT and Perplexity, and Google's AI Mode was found to cite 143% more unique domains than AI Overviews, which means newer surfaces are the more winnable ones for challenger brands.
6. The llms.txt question, settled
A file at your site root is not a visibility strategy. A study of 300,000 domains found 10.13% adoption, yet among the fifty most-cited domains in AI answers only one had published the file. Server-log analysis across more than 500 million AI bot visits over ninety days recorded a few hundred direct fetches. Google has stated on the record that it does not use the file for AI Overviews or AI Mode, and its 2026 guidance lists it among the things you do not need.
Publish one if you sell to developers or expect agent-driven lookups, because it costs half a day and carries option value. Do not let it displace the work that is measurably associated with citations: entity consistency across your own site and third-party profiles, visible authorship and dates, attributable claims, and coverage in the sources your category's engines already trust. Real crawler control still lives in robots.txt, where the major providers now run separate user agents for training, search indexing and live user fetches.
7. Search became a place to transact, not just to be found
The commercial layer moved fastest of all. Advertising went live inside ChatGPT on 9 February 2026 for free-tier US users, with a self-serve ads manager by May and rollout across six markets by July. In parallel, agent checkout standards shipped in production: a commerce protocol maintained by OpenAI and Stripe handles the agent-to-merchant checkout handshake, Google's Agent Payments Protocol handles cryptographically signed authorisation through intent, cart and payment mandates, and Google and Shopify's unified protocol became self-serve for developers on 17 June 2026.
eMarketer put AI platforms on track to process $20.9 billion of US retail spending in 2026, around 1.5% of US ecommerce and roughly four times the prior year. McKinsey's longer projection places agentic commerce at $3 to $5 trillion globally by 2030. For any brand selling online, the near-term work is unromantic: complete and accurate product feeds, resolvable identifiers, live inventory and pricing, machine-readable returns and shipping policy, and a checkout an agent can call.
8. What this means for the next two quarters
| Shift | Retire this | Run this instead |
|---|---|---|
| Impressions and clicks decoupled | Sessions as the sole visibility KPI | Track share of answers, citation counts and assisted revenue alongside sessions |
| Ranking no longer predicts citation | Single-keyword page targeting | Build topic clusters that answer the sub-queries a fan-out generates |
| Citations concentrate in few domains | Owned-media-only content plans | Earn presence in the community, reference and video sources your category's engines already cite |
| Comparison intent is over-indexed | Feature-list product pages | Publish honest comparison and alternatives pages with named, attributable sources |
| Search Console now isolates AI impressions | Third-party estimates as the only input | Baseline the new report now so a year of history exists when clicks are added |
| Agents transact, not just read | Feeds treated as a channel afterthought | Maintain feed accuracy, identifiers and policy data to the same standard as the site |
A measurement note worth acting on this month. Only around 14% of marketers currently track AI search performance at all, and an estimated 70% of AI-referred traffic can arrive without referrer data. Until that improves, the defensible approach combines three inputs: the Search Console generative AI report for Google surfaces, tagged and segmented referral traffic in analytics for the open-web engines, and a repeatable prompt panel of the fifty to one hundred questions your buyers actually ask, sampled on a fixed schedule across ChatGPT, Perplexity, Gemini and AI Mode.
Frequently asked questions
What is the difference between SEO, AEO and GEO?
SEO works to earn rankings and clicks in a list of links. Answer Engine Optimisation works to earn the direct answer position for a specific question. Generative Engine Optimisation works to earn citations and brand mentions inside AI-generated answers. The underlying work overlaps heavily: crawlable pages, clear structure, demonstrated expertise and a consistent entity footprint serve all three.
Has Google confirmed that GEO tactics work?
Google's position is that its generative features run on core Search systems, so no separate optimisation is required and no special markup or files are needed. It has not endorsed third-party GEO methods. That does not make GEO redundant, because ChatGPT, Perplexity and Copilot are not Google, and their citation behaviour differs materially from Google's.
How long does it take to build AI visibility?
Teams working systematically report meaningful movement in citation share over six to twelve months, with error correction in AI answers achievable within about two weeks once monitoring is in place. Brands with no monitoring typically discover citation losses and factual errors after two months or more of compounding.
Should we block AI crawlers?
For most commercial sites, no. Blocking removes the possibility of citation and referral without removing the competitive answer, which will simply be assembled from other sources. Publishers with licensing leverage face a different calculation.
See how AI engines currently describe your brand
LaCleo runs Generative Engine Optimisation and agentic SEO as a managed service. Multi-agent systems handle prompt monitoring, entity and schema review, comparison-content production and citation tracking, with human review on every deliverable. No dashboards to learn. We do the work.
Start with a free AI visibility audit: we run your category's buying prompts across the major engines and report where you appear, where competitors appear instead, and what is being said about you.
Run a free AI visibility audit or explore our SEO + GEO service.
Sources
- Google Search Central, Search Generative AI performance reports announcement, 3 June 2026, and Optimizing your website for generative AI features on Google Search, 15 May 2026 (last updated 10 July 2026).
- Google Search Central, AI Features and Your Website documentation.
- SparkToro, US zero-click search analysis, January to April 2026.
- BrightEdge, one-year AI Overview trigger analysis, 2026.
- Adobe Digital Insights, Q1 2026 AI referral conversion analysis.
- Ahrefs, 863,000-keyword AI Overview citation study, March 2026, and Brand Radar most-cited domains, June 2026.
- Semrush, 150,000 AI citations across 5,000 keywords, and 2026 cross-industry conversion analysis.
- Conductor, AI referral traffic benchmark across 13,770 domains.
- SE Ranking, ChatGPT referral traffic study (101,574 websites, May 2026) and llms.txt adoption study (300,000 domains).
- Digital Applied, 1,000 AI Overviews citation pattern study, April 2026.
- DeltaV Digital, 25,337-citation AI search citation study, April to July 2026.
- Tinuiti, Q1 2026 AI Citations Trends Report.
- eMarketer, US AI-platform retail spending forecast, December 2025; McKinsey QuantumBlack agentic commerce projection, October 2025.
- OpenAI and Stripe agentic commerce protocol documentation; Google Cloud Agent Payments Protocol announcement.