What Is Conversational Search?

Conversational search is searching by asking full questions and follow-ups in plain language, the way you would talk to a person, instead of typing a few keywords. For a health practice, it means some patients now describe their whole situation and refine it over several messages before they pick anyone to call. Google's AI Mode, for example, offers more advanced reasoning and lets the searcher keep asking follow-up questions.

How conversational search works

The old pattern was one search, one list of results, one click, then another search if the first did not work. The new pattern is a thread. The patient asks, the tool answers, and the next question builds on what was just said.

Three features make that possible.

• Plain-language input. The tool accepts a paragraph, not just keywords. A patient can write the whole story, including constraints such as "I don't want surgery" or "I'm on a budget."
• Memory inside the conversation. Perplexity's Help Center says it keeps the context of earlier questions, so follow-ups flow naturally. A patient can ask "Which of those offer telehealth?" and the tool knows what "those" refers to.
• Several searches behind each message. Google says AI Mode breaks a question into subtopics and issues many queries at once. That technique is called query fan-out, and each follow-up triggers a fresh round.

This happens in more than one product. Google AI Mode is the version inside Google Search, introduced in the U.S. on May 20, 2025. ChatGPT, with web search on, works as a conversation that can look things up. Perplexity is built around the same pattern, with sources shown for each answer.

What does this change for a practice? Consider how the shortlist forms. In a keyword search, a patient opens several tabs and compares. In a conversation, the comparing happens inside the tool. The first answer names a handful of options. The follow-ups narrow them. By the third message I expect the shortlist to be forming, and some patients will call from the answer without visiting a dozen sites. That is my inference from how the tools work.

That makes every follow-up a checkpoint. Each time the patient asks something, the tool has to find a source that answers it. A practice that has a clear page on the cost question, the telehealth question and the "what should I ask" question can be pulled into the answer at each step. A practice with a thin homepage and a "call us" button can only be pulled in once, at the start.

There is a technical requirement underneath, and it is unchanged. Google says a page must be indexed and eligible to be shown in Google Search with a snippet to appear in AI Overviews or AI Mode, and that there are no additional requirements. The conversation does not need new code from you. It needs pages that state facts.

This favors specific, honest content. Patients in a conversation are not looking for a slogan. They are comparing, and the tool looks for pages that state facts it can repeat. That is the logic behind answer-first writing, where each page opens with the answer and then adds detail.

It also favors completeness across the whole journey. Think of the patient's questions in three stages. First comes the problem, such as what is causing this and what are the options. Second comes the provider, such as who treats it, where and how. Third comes the decision, such as what does it cost, how long is the wait and how do I book. A common pattern is a blog that covers the first stage, an About page that touches the second, and little on the third. Conversational search tends to reach the third stage fast, so that is where gaps cost you.

One more wrinkle is that follow-ups often turn into comparisons. A patient may ask how you differ from a practice she has heard of, or whether a service is covered by insurance, and the tool may search for those sub-questions whether or not you have a page for them. Two patients asking the same opening question can also get different follow-up paths, because answers can be tailored to the person asking (see AI personalization). Your own test is one sample, not the whole picture. The safest coverage is a plain page or section for the comparison and money questions: your approach versus a common alternative, what the visit costs, and what insurance does or does not cover.

Example: conversational search at a health practice

This example is illustrative, not a transcript. A woman in Dallas opens a chat tool and types, "I'm 47, my sleep is terrible, my doctor says my labs are normal, and I want someone who actually looks at hormones. What should I ask?"

The tool answers with general guidance on perimenopause, a list of questions to ask a provider, and a note about which specialists treat hormone concerns. She follows up: "Do any of them do telehealth?" The tool searches again and gives options. She follows up once more: "What does it cost?"

At each step the tool needs a source. Consider two hormone clinics.

Clinic A has a homepage with a mission statement and a button reading "Contact us." Its services page lists "hormone optimization" in a bullet. It has no page on telehealth and no mention of cost.

Clinic B has a page titled "Hormone Therapy for Women in Dallas." It opens with two plain sentences on who the care suits and what the first visit includes. It has a short section titled "Do you offer telehealth?" that says yes and explains which visits can be remote. It has a section titled "What does it cost?" that explains how pricing works and gives a typical range for an initial visit. It lists the clinic address and provider credentials.

At the first step, either clinic might be named. At the telehealth step, only Clinic B has an answer to quote. At the cost step, only Clinic B has facts. By the third message, Clinic B is the one the tool can describe, and Clinic A is not.

Nothing about Clinic B is advanced. It anticipated the questions and answered them in a form a tool can lift. No one can promise it will be named, and the tool decides. But the odds shift toward the clinic that gave the tool something to say.

The practical step for Clinic A is to collect the questions its front desk and discovery calls hear most often, turn each into a heading with a direct answer on the relevant page, and publish a price range or a clear statement of how pricing works. That takes a few days, not months.

How to check yours in 60 seconds

1. Open ChatGPT, Perplexity or Google AI Mode.
2. Start with a long question from your ideal patient, including a constraint and your city.
3. Ask two follow-ups, one about cost and one about location or how to book.
4. See at which step your practice appears, if at all, and which pages are cited.

What to do about it

1. Collect the follow-up questions patients ask on discovery calls and at the front desk.
2. Give each its own heading and a two-sentence direct answer on a relevant page.
3. State your price ranges, or how pricing works, in plain words.
4. List your location, service area and telehealth availability clearly near the top of the page.
5. Add the clinician's name and credentials so the source is clear.
6. Add short sections for the comparison and money questions, such as "how this differs from [common alternative]" and "does insurance cover this."
7. Test a full three-message conversation each month and log which step you appear at. Use an invented patient story in your test, and keep real patient details out of it.