Skip to content
Answer engines6 min read

Conversational queries, and the shape an answer has to take

Typed search taught people to strip their questions down to keywords. Voice interfaces and assistants let them stop doing that. The queries are longer, more specific, more conditional — and a page optimised for a two-word phrase has nothing to say to them.

Part ofAnswer Engine Optimization (AEO)

For two decades people compressed their questions to suit the machine, because short keyword queries worked and full sentences did not. That constraint has gone. What arrives now is closer to what someone would actually ask a knowledgeable colleague, and it carries information the compressed version threw away.

What the longer query contains

Compressed queryConversational queryExtra information
multi tenant databaseshould each customer get their own database or share oneThe decision being made
dpdp consentdo we need consent to keep support tickets with customer namesThe specific case
nextjs slowwhy is my next js site slow on mobile but fine on desktopThe observed symptom and the contrast
seo agencyhow do I tell if an seo proposal is legitimateThe real anxiety behind the search

Each right-hand query names a situation. That is the opportunity: a page can address the situation directly, which a page targeting the left-hand phrase structurally cannot, because it does not know which of a dozen situations the visitor is in.

Four question shapes, and what each one needs

Conversational queries fall into a small number of shapes, and each wants a different response. Recognising which shape a question has is most of knowing how to answer it.

ShapeExampleWhat the answer must contain
Definitionalwhat is row level securityOne plain sentence, then the qualification
Comparativerow level security or separate databasesThe decision criteria, not a feature list
Suitabilityis row level security enough for healthcare dataAn honest conditional — when yes, when no
Proceduralhow do I add a policy without breaking queriesOrdered steps, with the failure mode at each

Comparative and suitability questions are where most pages fail, and they are where the commercial value sits. Both are asked by someone making a decision. Both are answered badly by a page that describes an option without saying when not to choose it.

Answers get follow-ups, so anticipate the second question

A conversational exchange does not end with the answer. There is a next question, it is predictable, and a page that answers it in the following section keeps the reader instead of returning them to the search box.

text
"should each customer get their own database or share one"
  -> "how do I stop one tenant reading another's rows"
     -> "what does that cost in query performance"
        -> "how do I move one tenant out later if they demand it"

Four sections of one page, in that order. Each one is also a
standalone answer to a query somebody types, and each one leads
to the next — which is the same structure as a good conversation.

This is the cheapest content-planning exercise available. Write the question, then write the question its answer provokes, three or four levels down. The result is a page outline that matches how the topic is actually thought about.

Spoken answers have constraints text does not

Where the answer is read aloud, several things that work visually stop working. A response delivered as speech is linear, unskimmable and unforgiving of length.

  • Short sentences. A sentence with three subordinate clauses is followable on a page and lost when heard once
  • No visual scaffolding. "See the table below" and "as shown above" have no meaning in speech
  • Numbers sparingly. One figure is retained; four in a sentence are not
  • Front-loaded. A spoken answer is often truncated, and always heard from the start
  • Plain words. A reader can reread an unfamiliar term; a listener cannot

None of this requires a separate voice version of the content. It requires that the first two or three sentences of each answer be simple, declarative and free of references to the page's own layout — which improves them for readers too.

Local and immediate intent

A meaningful share of spoken queries carry immediacy — near me, open now, today. Answering them requires facts stated unambiguously and consistently, not clever writing.

  • State the service area in words, not only by implication from an address
  • Keep name, location and contact details identical wherever they appear, including third-party listings
  • Do not assert a physical presence, service area or availability you cannot substantiate — an inaccurate claim here is corrected by the market quickly
  • Where availability genuinely varies, say what it is rather than leaving it to be inferred

What this does not mean

Two common overcorrections do more harm than the problem they address.

  1. Writing in fake dialogue. A page structured as an invented conversation between a customer and a salesperson is harder to read and no more extractable. Answer questions directly; do not perform a conversation.
  2. Stuffing question phrasings. Listing fifteen ways of asking the same thing, each with a two-line answer, is the keyword-stuffing habit in new clothing. One question, phrased as people actually ask it, answered properly.

The genuine change is smaller and harder than either shortcut: know the specific situations your audience is in, and answer each one with the conditions attached. That was always better writing. It is now also the thing that gets quoted.

Questions

How are conversational queries different from typed keywords?

They are longer, and the extra length carries information the compressed version discarded — the decision being made, the specific case, the observed symptom, the underlying concern. A page targeting a two-word phrase cannot address any of that, because it does not know which of a dozen situations the visitor is in.

Which question types are most worth answering?

Comparative and suitability questions — "X or Y" and "is X enough for Z". Both are asked by someone actively deciding, which is where the commercial value sits, and both are answered poorly by pages that describe an option without saying when not to choose it. The answer that gets cited names the conditions under which each choice is right.

Do I need a separate version of content for voice?

No. What spoken delivery requires is that the first two or three sentences of each answer be short, declarative and free of references to the page's own layout — no "as shown above", no sentence with three subordinate clauses, no four figures in a row. Those constraints improve the writing for readers as well.

Should I write content as a dialogue to match how people ask?

No. An invented conversation between a customer and a salesperson is harder to read and no more extractable than direct prose. The same applies to listing fifteen phrasings of one question with a two-line answer each — that is keyword stuffing in new clothing. Ask the question the way people ask it, once, and answer it properly.

Want this applied to your systems?

Most of what is described here begins as an assessment: a fixed-scope review producing a prioritised written report you can act on with any supplier.

Prefer email? hello@kornspire.com