The click is disappearing
68.01% of US Google searches ended without a click in early 2026, up from 60.45% in 2024 — the fastest acceleration in a decade.
Search has split into three surfaces — the results page, the direct answer, and the AI assistant that never shows a results page at all. We build visibility across all three, and we measure it in pipeline rather than in vanity rankings.
Every figure below is sourced and dated. Search behaviour is moving fast enough that an undated statistic is worthless — so we date ours.
68.01% of US Google searches ended without a click in early 2026, up from 60.45% in 2024 — the fastest acceleration in a decade.
Top-ranking pages see a 58% lower click-through rate when an AI Overview is present. Pew found clicks roughly halve, from 15% of visits to 8%.
51% of B2B software buyers now start research with an AI chatbot more often than Google — up from 29% a year earlier. 33% bought from a vendor they had never heard of.
AI platforms sent an average 770 million referral visits per month worldwide, up 117% year on year — and that traffic converts 54% better than non-AI sources in US retail.
They share a foundation and diverge in what they optimise for. Here is the honest comparison.
| SEO | AEO | GEO | |
|---|---|---|---|
| Full name | Search Engine Optimization | Answer Engine Optimization | Generative Engine Optimization |
| The goal | Rank high on the results page | Be the source chosen for the direct answer | Be retrieved, quoted and cited inside an AI-generated answer |
| Where it shows | Google, Bing organic listings | Featured snippets, People Also Ask, voice, AI Overviews | ChatGPT, Gemini, Perplexity, Copilot, Claude |
| What wins | Authority, relevance, technical health, links | Clear question–answer structure, schema, concise extractable claims | Citable evidence, entity clarity, presence across trusted third-party sources |
| Success metric | Position, impressions, organic sessions | Snippet and answer-box ownership | Share of voice in AI answers, citation rate, AI referral sessions |
| Content shape | Comprehensive pages built around a topic | Direct answers in the first 40–60 words, then depth | Statistics, quotable definitions, named sources, original data |
The disciplines are not rivals. Google states plainly that, from its perspective, optimising for generative AI search “is optimizing for the search experience, and thus still SEO” — and that a page must be indexed and eligible to appear with a snippet before it can be used in AI features. Technical SEO is the entry ticket; AEO and GEO decide what happens once you are through the door.
Crawlability, index coverage, Core Web Vitals, rendering, canonicalisation, sitemaps and log-file analysis. Nothing downstream works if crawlers cannot reach or render your pages.
Schema.org markup for organisation, service, product, article, FAQ and review — so machines understand not just your words but what they refer to.
Mapping the questions your buyers actually ask across the whole journey, then structuring the site so each one has an obvious home.
Pages written to answer the question in the first 40 to 60 words, then go deep — the structure both featured snippets and AI retrieval reward.
Citable statistics, clear definitions, original data and named sources, so an AI model has something concrete to quote and attribute to you.
Digital PR, directory and comparison-site accuracy, and earned mentions — because models draw on what the wider web says about you, not only your own site.
Tracking how ChatGPT, Gemini, Perplexity and AI Overviews respond to a defined prompt set for your category, and how often you appear.
A dashboard tying rankings, AI citations, referral sessions and branded search back to pipeline — not a PDF of impressions.
Most agencies still sell a keyword report and a monthly blog post. That was adequate when the results page was the whole game. It is not any more. Our programme runs on a 90-day cycle with a clear objective for each phase and a dashboard you can open whenever you like.
Technical crawl, content and schema review, competitor and AI-answer benchmarking. You get the findings as a document you own.
Technical remediation, structured data rollout, information architecture and internal linking corrected.
Answer-ready content against the priority question set, plus digital PR and third-party accuracy work.
AI share-of-voice tracking, ranking movement, referral analysis and a prioritised plan for the next cycle.
AI visibility is newer than SEO, which means the industry is full of confident claims and thin evidence. We report what is actually measurable, name the method, and say plainly when something is directional rather than precise.
Where buyers research before they ever speak to sales. G2 found 69% of buyers selected a different vendor than planned based on chatbot guidance, and 33% bought from a vendor they had not previously heard of. If the AI does not know you, you are not on the list.
Where product discovery increasingly starts in an assistant rather than a search box, and where AI-sourced traffic has been converting markedly better than other channels.
Where a handful of high-intent enquiries per month decides the year, and being the answer to “who is the best X in Y” is worth more than a thousand impressions.
No — but the job description changed. Google's own documentation states that a page must be indexed and eligible to appear in Search with a snippet before it can be used in AI features, and that optimising for generative AI search “is optimizing for the search experience, and thus still SEO”. What has died is the assumption that a top ranking automatically delivers the click. That is what AEO and GEO exist to address.
Google explicitly says no — it does not require llms.txt, content chunking, or a special AI writing style. Be sceptical of anyone selling those as essential. What genuinely helps is accurate structured data, clear factual claims and a site that renders properly without JavaScript gymnastics.
Technical fixes and structured data can change eligibility within weeks. Movement in AI citations and rankings typically takes three to six months, with compounding returns after month six. Anyone promising you first-page results in 30 days is either misleading you or planning to do something that will hurt you later.
We run a fixed prompt set — the real questions your buyers ask — across the major assistants on a scheduled cadence, and record whether you appear, whether you are cited as a source, and where you sit relative to competitors. It is sampling, not a complete census, and we say so in every report.
No, and neither can anyone else honestly. Generative answers are probabilistic and change between model versions. What we can do is make you the most citable, best-structured, most corroborated source in your category — which is what demonstrably raises the odds.
It can be either. Folding it into a build is cheaper and better, because technical SEO, structured data and content architecture get done correctly at launch rather than retrofitted a year later. As a standalone programme it works on any site we can access.
Use AI to research and draft, then have a human with actual expertise verify, sharpen and add something the model could not know. Published at volume without that step, AI content tends to be generic, unverifiable and precisely the kind of source a model has no reason to cite.
We will run your category through the major assistants, audit your technical and structured-data foundation, and show you exactly where you stand — before you commit to anything.