TL;DR — the one-paragraph answer
In 2025, AI search broke the classical sales funnel in two directions at once. Organic CTR on queries with a Google AI Overview collapsed 61%, from 1.76% to 0.61%. At the same time, the average AI-search visitor became 4.4× more valuable by conversion than a traditional organic visitor. Fewer clicks, but each one is worth far more — if the AI cites your brand. Getting cited is a separate discipline called GEO, and it no longer follows from strong classical SEO.
What is GEO (generative engine optimization)?
GEO (generative engine optimization) is the practice of structuring a website and its content so that AI systems — ChatGPT, Perplexity, Google AI Overviews, Gemini, and Copilot — can read, extract, and cite it as a source in their generated answers. Where classical SEO optimizes to rank a page in a list of blue links, GEO optimizes to get the page quoted inside an AI-written answer.
The two use different signals: GEO relies on server-side rendering, machine-readable Schema.org markup, an llms.txt file, self-contained answer passages, and external citations — not keyword density and backlinks alone. A 2024 paper from Princeton and the Allen Institute for AI found that correctly preparing content for generative engines yields up to a 40% lift in visibility inside AI answers.
A note from me before the numbers
Over the last two years I've gone through dozens of studies on how AI search is reshaping user behavior and the sales funnel — from Ahrefs and Semrush to Princeton, McKinsey, and Bain. In parallel I've been rebuilding my own site for that new reality. This article is a short distillation: the one core contradiction of 2025, the one structural change in how clients now make purchase decisions, and a clear answer to "do I need to do something about this right now?" No loud forecasts — only numbers from primary sources.
How far did CTR actually fall in 2025?
Organic CTR fell by roughly 60% on AI-influenced queries over 2025 — the single largest two-year drop on record.
In December 2025, Ahrefs published a study across 300,000 queries: CTR for the top organic position in Google dropped 58% over two years. In November, Seer Interactive analyzed 3,119 queries and 25 million impressions and landed in the same place — on queries with an AI Overview, organic CTR collapsed from 1.76% to 0.61%, a 61% drop. Paid CTR fell too: 19.7% to 6.34%.
In parallel, Semrush studied 500+ topics and recorded the opposite finding: "the average AI search visitor is 4.4 times as valuable as the average visit from traditional organic search, based on conversion rate." This isn't a contradiction. This is the main story of 2025.
What actually happened to user behavior?
AI assistants moved from a curious minority to default infrastructure in under two years.
By early 2026, AI assistants are no longer a niche tool — they're a mass channel where people make decisions. ChatGPT has 900 million weekly active users and 2.5 billion queries a day; Google AI Overviews reach 2.5 billion monthly users; Google AI Mode hit 1 billion monthly users a year after launch (Reuters, Alphabet, Adobe, Similarweb). The key point: this audience barely clicks out. 60% of regular Google searches already end without a visit to any site, and in AI modes that reaches 93% — the user reads the ready answer and closes the tab. Put the two facts together and the conclusion follows: a growing share of your potential clients now decide inside the AI answer, without ever opening your site. That's why the classical funnel — saw in search → clicked → arrived on site → bought — is physically falling apart.
How does a client's buying decision now work?
The decision point has moved upstream — from the comparison page on your site to the AI answer the client reads before ever visiting your site.
It used to be: a client googled, clicked ten blue links, visited five sites, compared. Now they ask ChatGPT, "what are the best platforms for X right now?" and get back three to five names. If your brand is in that list, a micro-sale has already happened. If it isn't, for that client you don't exist.
Classical funnel (pre-2024)
AI-search funnel (2025→)
Bain & Company has tracked this for years: 85% of B2B buyers arrive with a "Day One List," and in 95% of cases the winner comes from that list. Increasingly it's built from an AI chat answer. G2's March 2026 study of 1,076 B2B buyers found 51% now begin research inside an AI chatbot, and 69% ended up choosing a different vendor than originally planned.
Two consequences follow. First: traffic to your site will drop, and that's the normal new state. Second: the traffic that's left will be worth several times more, because the people arriving have already made a preliminary decision. That's why Semrush's 4.4× exists, why Adobe shows +254% revenue per visit, and why Visibility Labs records ChatGPT conversion of 1.81% versus 1.39% for non-branded organic.
Why doesn't strong SEO guarantee AI citations anymore?
Strong Google rankings no longer predict AI citations, because AI engines now cite mostly different sources than the classical search top.
BrightEdge reports that the overlap between AI Overview citations and Google's top 10 organic results dropped from 75% in mid-2025 to 17–38% by early 2026. The Princeton/Allen Institute "GEO" paper (2024) shows correctly prepared content yields up to 40% more AI visibility. A separate University of Toronto study (2025) shows AI engines cite earned media — external mentions and third-party publications — roughly 5× more often than content on the brand's own site.
So the new model requires three things at once: (1) make the site technically readable for AI crawlers, (2) structure content the way LLMs expect to receive it, (3) systematically appear in sources the AI considers worth citing. Classical SEO covers only part of (1). That gap is what GEO fills.
Who has already won and who has already lost?
The redistribution of visibility from AI search has already happened: sites dependent on classical search clicks lost traffic and revenue, while sites the AI began citing gained both. This is not a forecast — it is a 2025 fact.
| Losers (classical-click-dependent) | Winners (AI-cited) |
|---|---|
| Chegg — revenue −36%, non-subscriber traffic −49%, market cap −99%, suing Google | Walmart — ~20% of referral traffic now from ChatGPT |
| Business Insider — organic search −55%, 21% staff cut | Etsy — >20% referral from AI, stock +16% on Instant Checkout news |
| Stack Overflow — 75% drop in new questions from peak | Reddit — +450% AI citations |
| Penske Media (Rolling Stone, Variety) — over ⅓ of affiliate revenue gone in 2024 | Tripadvisor & Expedia — partnerships with OpenAI and Perplexity |
Gartner predicted back in February 2024 that classical search volume would drop 25% by 2026. We are inside that prediction now.
What does a site need to do to land in the AI answer?
The basic GEO foundation comes together in a single sprint of six technical moves; the harder, higher-leverage layer is earned media.
Server-side rendering
So the AI crawler sees content, not a blank page. The most important step — without it the other five are useless.
Open access to AI bots
GPTBot, ClaudeBot, PerplexityBot and others, in robots.txt and at the server level. Permission is useless if a firewall blocks the bot at the server.
Machine-readable Schema.org markup
Person, Service, FAQPage, Article — plus a separate llms.txt file, the "table of contents for language models."
Correct bilingual handling via hreflang
So AI doesn't treat language versions as duplicates and serves the right one to each audience.
Page speed in the green of Core Web Vitals
Slow pages get skipped by bots to save crawl budget. Speed is a visibility factor, not cosmetics.
IndexNow plus webmaster submission
Google Search Console, Yandex Webmaster, Bing Webmaster. Without IndexNow a new page indexes in weeks; with it, in minutes.
That's the work on your own territory. On top of it sits earned media — systematically appearing in sources AI engines trust. It's longer and more expensive, but it's the lever behind the 5× citation lift the Toronto paper documents.
How I did this on my own site — a six-step walkthroughWhy order GEO from me?
Because the method is already proven on a verifiable, public asset — this very site — and on client projects, not on slideware.
Two cases sit on this blog walking through, in first person, how I built my own site as proof of method and how I then rebuilt it for AI search. The same method runs on client projects. Not "10 years in SEO," not "a team of 50 experts" — a specific site, specific steps, specific results you can verify right now: ask ChatGPT or Perplexity "who is Roman Denisov, AI architect of revenue."
This is the exact principle that broke classical SEO marketing: AI doesn't cite the loudest sites, it cites the most verifiable ones. If the method works on my own site, it's applicable. If I were writing this article about GEO while my own site sat on a stale drag-and-drop builder with a "contact us" form, this would be empty advertising — and the post wouldn't exist.
Want to know whether AI can see your site?
I run a free express diagnostic. The follow-up is a 30-minute call where I show you, side by side, how a person currently sees your site and how a neural network sees it — plus the concrete list of what to change for AI engines to start citing your brand.
Key numbers and sources
| Metric | Value | Source |
|---|---|---|
| Top-1 organic CTR change (2 years) | −58% | Ahrefs, 300k queries (Dec 2025) |
| Organic CTR on AI-Overview queries | 1.76% → 0.61% | Seer Interactive (Nov 2025) |
| Value of an AI-search visitor vs organic | 4.4× | Semrush, 500+ topics |
| Revenue per visit from AI traffic | +254% YoY | Adobe |
| ChatGPT vs non-branded organic conversion | 1.81% vs 1.39% | Visibility Labs |
| Searches ending without a click (AI modes) | 93% | Similarweb / industry |
| AI-Overview vs Google top-10 source overlap | 75% → 17–38% | BrightEdge |
| Visibility lift from GEO-optimized content | up to +40% | Princeton & Allen Institute for AI (2024) |
| Earned-media citation rate vs own-site content | ~5× | University of Toronto (2025) |
Sources
- •Ahrefs — CTR study across 300,000 queries (December 2025)
- •Seer Interactive — 3,119 queries / 25M impressions, AI Overview CTR analysis (November 2025)
- •Semrush — AI search visitor value study, 500+ topics
- •Adobe — AI-driven retail traffic and revenue-per-visit data (2025 holiday season)
- •Visibility Labs — ChatGPT vs organic conversion data
- •Similarweb / Reuters / Alphabet — AI assistant usage figures (early 2026)
- •Bain & Company — B2B "Day One List" buyer research
- •G2 — B2B buyer behavior study, 1,076 buyers (March 2026)
- •BrightEdge — AI Overview vs organic top-10 source overlap
- •Princeton University & Allen Institute for AI — "GEO: Generative Engine Optimization" (2024)
- •University of Toronto — earned-media AI citation study (2025)
- •Gartner — search volume forecast (February 2024)
