Alvaro Peña de Luna 
Conversational search is a way of searching in which you ask questions in natural language and refine them through follow-up messages, while the engine keeps the context of the conversation and answers in prose instead of a list of links.
How conversational search works
In keyword search, each query stands alone. You type "crm small agency", scan ten links, then type a new query if you need more. In conversational search, the engine remembers what came before:
- You ask a full question: "What CRM would you recommend for a 10-person marketing agency?"
- The engine answers with a few options, often explaining who each one suits and citing sources.
- You follow up: "Which of those has the best client reporting?"
- The engine narrows the list using the context of the first answer.
Behind the answer, a large language model interprets the question, often runs web searches, and writes a reply from what it found. ChatGPT, Perplexity and Google AI Mode are the best-known examples.
Conversational search vs keyword search
| Keyword search | Conversational search | |
|---|---|---|
| Input | A few words | A full question with context |
| Memory | Each query is new | Follow-ups build on earlier turns |
| Output | A ranked list of links | A written answer, sometimes with sources |
| Where brands compete | Position on the page | Being named, and how |
Why conversational search matters for brands
A conversation tends to end in a shortlist. After two or three follow-ups, the user has a small set of brands that fit their situation, and the others never come up again. If you are not in the first answer, a follow-up rarely brings you back.
Answers are also more nuanced than rankings. An engine can recommend you "for large teams" and a competitor "for freelancers", or mention you as an example without recommending you. The way you are named matters as much as whether you are named, which is why metrics like AI brand sentiment exist.
How to optimize for conversational search
- Answer real questions. Write pages and sections around the questions buyers ask, in their words, including comparisons and "which is best for…" questions.
- Make your fit explicit. State who your product is for, what it costs and where it falls short. Engines reuse clear statements when they qualify a recommendation.
- Cover the follow-ups. Pricing, integrations, alternatives and limitations are the questions that come next in the conversation.
- Get mentioned in third-party sources. Reviews and comparisons give engines independent evidence to name you.
How to measure it
Conversations are private, so you measure with a representative set of prompts and repeat the checks over time:
Mention rate = answers that name your brand / answers checked × 100
Worked example: 40 prompts checked on 3 engines once a week for a month give 480 answers. If your brand is named in 120, your mention rate is 25%. Split it by engine and by prompt group to find where conversations leave you out. Our guide to AI search optimization covers what to do with the results.
Common mistakes
- Tracking two-word keywords when users type full questions.
- Testing a prompt once in your own logged-in account and treating the answer as typical.
- Counting mentions without checking whether they are recommendations or passing references.
How Mencoro tracks conversational search
Mencoro runs the prompts you choose on ChatGPT, Perplexity, Google AI Overview and Google AI Mode, in the countries you pick, with 1 to 3 passes per check to smooth out answers that vary. Each check is a single question with no account, history or personal context, so it shows how the engine answers the question itself rather than a follow-up in someone's conversation. Every brand named is classified by type (recommendation, comparison, listing, example or reference), tone, and whether the mention is direct or conditional ("for large teams"), which feeds Coverage, Favorability and Share of voice. See the AI brand monitoring feature, the methodology page, or test a prompt with the free ChatGPT rank checker.
Related glossary terms
- Google AI Mode: Google's conversational search tab.
- Answer engine: a system that replies with an answer instead of a list of links.
- Prompt tracking: checking how AI engines answer a set of prompts over time.
- Large language model (LLM): the model that writes the answers.
