Engines assemble recommendations from evidence, and not all evidence is equal. Six types do most of the work: specifications, pricing, fit statements, third-party corroboration, recency signals and exclusions. Knowing which type access which prompt family turns vague "improve the content" advice into a scheduled workstream.
Specifications serve comparison prompts. When a buyer asks which of two products handles a given requirement, the engine needs numbers on both sides. Missing specs do not produce a cautious mention, they produce silence, because the engine simply cannot complete the comparison with you in it.
Pricing opens budget-constrained prompts, which are a large share of all commercial prompts. Published pricing, even as a range or a starting point, outperforms "contact us" dramatically in answer engines. If the client cannot publish exact prices, publish bands and the factors that move them.
Fit statements serve situational prompts: who this is for, at what size, in what circumstances. These are the statements your hints mirror, and they must exist in public material, not only in the hint. A hint is a targeting instruction; the fit statement is the evidence that backs it.
Third-party corroboration opens trust-sensitive prompts. Reviews, analyst mentions, customer stories with named companies and directories that repeat your specs all raise confidence. Self-published claims echoed nowhere else are treated as weaker, particularly in regulated or high-consideration categories.
Recency and exclusions are the two most neglected. Dated, regularly updated pages are increasingly favoured over undated ones, so a visible last-reviewed date is cheap and effective. And explicit exclusions, what you are not for, increase trust measurably, because they make the rest of your claims read as calibrated rather than promotional.