Generative Search
Generative search is when a system answers a query with a written response synthesized from multiple sources, instead of a ranked list of links to click through. Google AI Overviews, ChatGPT, Gemini, Perplexity and Copilot all work this way for at least some queries. Your page can rank well and still be invisible here — what counts is whether the model pulls your content into the answer it writes, or names you as a source.
Why this matters for you
Generative search changes what a click is worth. Pew Research, working from real browsing data on roughly 68,879 Google searches, found users clicked through to a traditional result in only 8% of visits where an AI summary appeared, versus 15% without one — about half the rate. Ahrefs separately found the #1 organic result sees a 58% lower average click-through rate on queries where an AI Overview shows up. Zero-click behavior is now the norm, not the exception: SparkToro and Similarweb put 68% of US Google searches at zero clicks to any website in early 2026, up from 60% two years earlier. None of this means traffic is gone; it means the moment where you win or lose attention has moved earlier, into the answer itself.
How it works technically
Most generative search systems run on retrieval-augmented generation: a retrieval step pulls candidate passages from an index or live crawl, then a language model writes an answer grounded in what it retrieved. This is why structure and clarity matter more than keyword density — a passage has to be extractable and self-contained to survive that first step. It's also why Google has been explicit that AI Overviews and AI Mode need no special markup: no dedicated schema, no llms.txt, no content written separately "for AI." John Mueller confirmed in 2025 that no Google Search system reads llms.txt at all, and an Ahrefs analysis of roughly 137,000 sites that published one found about 97% saw no measurable referral traffic from it. Write for the ordinary reader and the retrieval step first.
Common mistakes
The biggest mistake is treating generative search as a markup problem. Ahrefs' research shows only 6–8% of URLs ChatGPT cites overlap with Google's top 10 for the same query, and around 80% of ChatGPT's cited URLs don't rank in Google's top 100 at all — citation and ranking are different selection processes, so optimizing purely for one leaves the other untouched. A second mistake is chasing links while ignoring mentions: an Ahrefs study of about 75,000 brands found mention frequency correlates with AI citation rate around 0.664, roughly three times the correlation for backlinks (about 0.218). Third, accuracy isn't optional theater — the Columbia Journalism Review's Tow Center found AI search tools misidentified the source of a news article more than 60% of the time across 1,600 test queries, so vague or unverifiable claims on your own pages give a model every reason to skip you or get you wrong.
Relation to AI recommendations
Generative search is where an AI recommendation actually gets made. Ask an assistant which tool, vendor or service fits a need, and it drafts a short list from whatever it judged citable and credible during retrieval. Getting into that list is the practical goal of generative engine optimization, the discipline named and studied in the 2024 ACM SIGKDD paper "GEO: Generative Engine Optimization" (Aggarwal, Murahari, Narasimhan, Deshpande, Rajpurohit and Kalyan). With ChatGPT at 900 million weekly users, Gemini past 1 billion monthly users, and Google AI Overviews reaching over 2 billion monthly users across 200+ countries, the volume of queries resolved this way is no longer a niche case — it's a growing share of how people find a business at all.
Example
Someone asks an AI assistant, "which invoicing software works for a two-person design studio?" Instead of ten blue links, they get a short written answer naming two or three tools, each with a line of reasoning and sometimes a source link. A small bookkeeping tool with a clear, well-structured comparison page might get named; a competitor with a stronger product but a page buried behind a signup form or a wall of marketing copy might not. The user reads the answer, picks one, and never sees a results page at all. That's the moment generative search decides — well before what used to count as a website visit.
Common questions
Is generative search the same thing as an AI Overview?
AI Overviews are Google's implementation of generative search, sitting above the normal results. ChatGPT, Perplexity, Gemini and Copilot are separate generative search systems with their own retrieval and source-selection logic, so being cited by one doesn't guarantee being cited by another.
Do I need special markup like schema or an llms.txt file?
No. Google's own guidance says no special markup is required for AI Overviews or AI Mode, and Google has confirmed no Search system reads llms.txt. Clear, accurate, well-structured content that a retrieval system can extract cleanly matters far more than any dedicated AI file.