Content built from the question, not the keyword

The brief starts with the answer gap

A keyword list tells you what people search. It does not tell you which page the engine cited instead of yours, which claim was missing, or which competitor got named in the paragraph where you should have appeared. Sonarvue begins with the answer. The brief records the prompt, the answer, the cited sources, and the specific gap the page has to close.

That changes the work. Instead of another generic article, you get a page with a job: answer the question directly, supply the proof the model needs, and make the comparison legible to a buyer who is already deep in research.

Structure for extraction

• Answer first. A direct response in the opening paragraph, before the history and the caveats.
• Evidence where it matters. Numbers, sources, definitions, and named entities that a model can quote without guessing.
• Comparison and boundaries. Who the product is for, where it stops, and how it differs from the alternative the buyer is weighing.
• Machine-readable structure. Headings, lists, tables, and schema that match the question the buyer asked.

Your voice, your approval

The agent can draft from your existing pages, product facts, and approved claims. It flags any claim that needs a source and leaves the final wording with you. Nothing publishes until a person approves it. If you already have a writer, the agent hands over the brief and the evidence instead of the draft.

Measure the answer, not word count

After publication, the same prompt runs again. You see whether the page entered the answer, whether it became a citation, and whether the competitor set moved. If the answer does not move, the next brief starts from what the engine did, not from a content calendar.