Is AI search eating SEO? What to do while competitors are still thinking
Up to 30% of B2B queries now start in ChatGPT and Perplexity. Based on data from 40 of our projects: how traffic is shifting, which pages AI engines cite, and what to implement first — structure, Schema.org, facts and brand citability.
In 2024 the question was “will AI search matter?”. In 2026 the numbers answer it: across 40 client projects we track, AI assistants and AI Overviews already influence 18–34% of commercial journeys depending on the niche. B2B feels it first — up to a third of decision-makers ask ChatGPT or Perplexity for vendor shortlists before they ever touch Google.
What actually changed in the traffic
Classic organic clicks are not collapsing — they are being filtered. Informational queries increasingly end inside the AI answer (zero-click), while commercial and local queries still send clicks, but fewer and warmer ones. The visitors who do arrive from AI answers convert 1.5–2× better in our data: the assistant has already pre-sold them.
- Informational content: expect fewer clicks, plan for citations instead;
- Commercial pages: clicks remain, but the shortlist is formed in the AI answer;
- Brand queries: growing — people verify what the AI recommended.
Which pages AI engines actually cite
We reverse-engineered citations across ChatGPT Search, Perplexity and AI Overviews for client niches. The cited pages share a pattern: they answer one question early and completely, carry verifiable facts (numbers, dates, names), and are structurally boring — clear headings, short definition blocks, FAQ markup. Beautiful storytelling landing pages almost never get quoted.
- A direct answer in the first 2–3 sentences of a section;
- Facts with numbers and sources — LLMs prefer attributable claims;
- FAQPage / Organization / Product Schema.org markup;
- Consistent brand facts across the web: directories, reviews, profiles.
The playbook: what to implement first
First, fix access: check robots.txt for GPTBot, PerplexityBot, Google-Extended and add an llms.txt with your key pages. Second, restructure money pages: add a “quick answers” block with 3–5 question-formatted facts. Third, build the brand corpus — profiles and mentions on sources models retrieve: industry directories, review platforms, comparison articles. Fourth, measure: run a fixed panel of 20–50 niche queries against the main AI engines monthly and record who gets named.
GEO doesn’t replace SEO — it’s the same foundation with a second consumer. Google reads your structure, and so does ChatGPT. Build once, get cited twice.
The window matters: in every niche we track, the first brand to optimize becomes the “default answer” and holds it — models reinforce what they already cite. Doing this six months before competitors costs less than displacing them after. If you want the numbers for your niche, request our free AI visibility check: we’ll show who ChatGPT recommends today for your money queries — you or someone else.