Answer Radar
How Answer Radar Proposes Buying Questions
Pressing Generate questions asks Answer Radar to write commercial-intent buying questions for your project. They arrive as a draft set, unapproved, for you to keep or cut. This article explains what the proposal uses, what it refuses to do, and the two ways to check a proposed set against what buyers really type.
Quick Answer
Generate questions creates a draft set of proposed commercial-intent buying questions from your project's domain, brand, competitors, category and buyer segment. Nothing is measured until you approve. Regenerate for another set, and check proposals against Search Console queries where connected.
- Proposals are a starting point, not the prompt set. Approval is yours.
- Types are not generated. You classify a question's shape and funnel stage yourself.
- Search Console queries, once connected, are the sanity check the proposal cannot be.
On this page
What the proposal draws on
The proposer reads the project's commercial context: the domain, the brand name, the competitors you named, the category and the buyer segment. From that it writes questions a buyer with commercial intent would ask an assistant, such as best tools for the job, alternatives to a named competitor, comparisons and pricing questions. Through MCP the same capability is answer_prompts_propose, which takes a project id and a count.
The result is a draft prompt set. Every question in it is unapproved, and the set stays draft until you approve one, at which point it becomes active. Regenerate produces another draft set; nothing already approved is touched.
What the proposal deliberately does not do
It does not approve anything. Deciding what gets measured is a separate act from suggesting candidates, and the product keeps them apart so a generated list never quietly becomes your spend.
It does not classify. Type, one of best_of, comparison, alternatives, pricing, how_to or review, and intent such as bofu are labels you assign after reading the questions. Nothing generates a type for you.
It does not estimate prompt volume. No engine publishes how often a private assistant is asked a question, and the proposer will not invent a number to rank suggestions by. The methodology page says the same about every number in the product.
Grounding proposals in Search Console
If the project is connected to Google Search Console, the gsc_queries_get tool returns the queries real buyers used to reach your site, ranked by impressions, with clicks and average position. That is first-party demand rather than a guess, and it is the best test of a proposed set: which proposals match how people actually type, which high-impression questions you rank poorly for and should measure, and which proposals no one asks.
Add the queries worth measuring with 'Add your own buying question', phrased as a buyer would ask an assistant rather than as a search string. The connection itself takes a few minutes and is covered in the Search Console article.
A working routine
Generate once. Read every proposal and ask whether a real buyer would type it before choosing. Approve those, remove the rest. Add the questions you already know from sales calls and support tickets. Classify by type so every type is covered, tag by product line or persona, and only then run the first audit.
Re-run the routine when you launch a product line or a competitor changes positioning, not weekly. A stable prompt set is what makes the series comparable over time.
FAQ
Why are some proposals generic?
Because the project's category or buyer segment was generic. The proposer only knows what the form told it. Tighten the category and segment, or write the specific questions yourself; your own wording is always the better source.
Does regenerating cost credits?
Generating proposals does not measure anything, so it does not spend measurement credits. Credits are spent only when an audit asks an engine a question.
Can proposals include my brand name?
They should not, and you should remove any that do. A question that names you measures whether the engine recalls you, not whether it recommends you among alternatives, and recommendation is what the gaps are about.
Related help pages
Answer Radar
Choosing and Approving Buying Questions
How the Questions page works: generate, approve, remove, add your own, classify by type and funnel stage, tag with topics. Only approved questions are audited.
Answer Radar
Connecting Google Search Console
The service-account flow: add Linkeddit's address as a user on your Search Console property, check for properties, pick one. No OAuth, read-only, and what.
Answer Radar
Segments and Topics: How Results Are Grouped
Why every metric is reported per segment and never averaged across engines, what the five axes are, what the instrument token is, and how topics group prompts.
Answer Radar
What Answer Radar Does Not Do
The honest list: no prompt volume, no traffic attribution, no causal claims, no consumer-app reproduction, no written briefs, no publishing, no model ids,.