Conversational Search
Conversational search means asking a question in plain language and refining it through back-and-forth, instead of typing isolated keywords into a search box. The system holds context across turns, so a follow-up like "what about something cheaper" still makes sense without you repeating yourself. ChatGPT, Perplexity, Google's AI Overviews, and Gemini all work this way. The result you get is usually a direct answer, not a list of links to click through.
Why this matters for your visibility
With a classic query like "hotel Vienna city center cheap" you get ten blue links and pick one. With conversational search you ask "where should I stay in Vienna near the opera for under 120 euros" and the assistant just answers, often with a named recommendation and no click required. Google has said AI Overviews now reach over 2 billion people a month across 200+ countries, and Gemini alone passed 1 billion monthly active users in August 2026, so this isn't a fringe behavior anymore. Pew Research, working from real browsing data on US adults, found people clicked through to a traditional result in only 8% of visits when an AI summary appeared, versus 15% when it didn't — roughly half the click rate. Visibility increasingly means being the name the model says out loud, not the link with the best position.
How it works technically
Conversational search runs on a large language model that predicts language from patterns learned in training, plus a context window that holds the running conversation so it can track what you already asked. Many systems pair that model with retrieval-augmented generation: before answering, they pull in current web pages and ground the response in what they find, rather than relying only on what was baked into the model during training. That's good news for site owners, but it comes with a catch. Google has stated plainly that no special markup, schema, or AI-specific file is required for a page to be picked up by AI Overviews or AI Mode, and that writing content specifically "for AI" is the wrong instinct. What actually helps is content a retrieval system can parse cleanly: real text rather than an image of the answer, and facts that don't contradict themselves from one page to the next.
Common mistakes
The most common mistake is optimizing only for short keywords when conversational queries are longer and phrased as full questions; a page tuned for "cheap hotel Vienna" can miss the actual question being asked. The second is chasing AI-specific files as a shortcut: an Ahrefs analysis of roughly 137,000 sites that published an llms.txt file found about 97% saw no measurable referral traffic tied to it, and Google's John Mueller has confirmed no Google Search system reads these files at all. The third is hiding facts — prices, hours, availability — inside images, PDFs, or menus a retrieval step can't read cleanly, and letting the same fact appear differently across pages, which makes a brand look unreliable to a system that's trying to decide what to trust and repeat.
Relation to AI recommendations
A conversational search rarely ends with a list; it ends with a recommendation, something closer to "I'd suggest X, because…" than "here are your options." That single-answer format raises the stakes: if you aren't named, you don't get considered. It also explains why citation and ranking behave so differently here than in traditional search — Ahrefs found only about 6–8% of URLs cited by ChatGPT overlap with Google's top 10 for the same query, and roughly 80% of ChatGPT's cited URLs don't even rank in Google's top 100. Brand mentions across the web turn out to matter more here than links: the same Ahrefs research found mention frequency correlates with AI citation rate at about 0.664, roughly three times stronger than the correlation for backlinks. Tracking how often and how favorably you show up, sometimes called mention rate or share of voice, is how you measure progress in this format.
Example
Say someone is planning a birthday dinner and asks an assistant: "I need an Italian restaurant in Cologne for Saturday night, eight people, with vegan options and a quiet side room." Instead of a results page, they get two or three concrete suggestions with reasons attached. A restaurant that spells out its menu, seating capacity, and vegan dishes in plain text on its own site is easy to recommend. One that only has that information as a photo of a printed menu is invisible to the same query, even though it would have been a perfect fit. The difference decides who gets the table on Saturday.
Common questions
Is conversational search the same as voice search?
No. Voice search just describes the input method: speaking instead of typing. Conversational search describes the dialogue itself: the ability to ask a follow-up, refine an earlier question, and stay in context over several turns, whether you typed or spoke it. The two overlap often, since voice interfaces are usually conversational, but they aren't the same thing.
Do I need an llms.txt file or special AI schema to show up?
No. Google has said explicitly that no special markup, schema, or AI-specific file is required for AI Overviews or AI Mode, and Ahrefs found essentially no traffic benefit across the tens of thousands of sites that added an llms.txt anyway. What actually helps is content that's legible to a retrieval system: real text instead of images, unambiguous facts, and consistency across your own pages.