Define specific questions
Create a set of questions a buyer might ask about your category. Include comparison, selection and requirements, not only questions naming your brand. Record language, system, date and exact wording. Without those conditions, two screenshots may look comparable while representing different contexts.
Classify what appears
Distinguish a brand mention, a link, an accurate description and a recommendation. Check for services or facts incorrectly attributed to your business. Retain enough evidence to review the classification. Positive visibility may be useful, but it does not on its own prove that a visitor chose your business.
Improve verifiable information
Keep services, terms, authorship and sources consistent. Explain processes and limits through examples a reader can inspect. Improvements should also help someone visiting the website directly. Do not base the proposition on magic formulas or a promise to control how an external system will answer.
Connect visibility and demand
Review available referrals, incoming questions and commercial quality while acknowledging attribution limits. Repeat the question set under documented conditions and retain variations. Report trends and observations, not an invented share of the AI market derived from a handful of queries.
Worked example: a small measurement protocol
For a hypothetical initial test, define 12 questions: four about supplier selection, four about service scope and four about buyer problems. Keep the wording and avoid inserting the brand into every question. Run the protocol in agreed systems with language and date recorded. For every answer, note mentions, links, accurate descriptions and errors. Do not turn three mentions in twelve questions into market share: the sample was selected for learning, not to represent every buyer. Repeat under comparable conditions and examine changes alongside published content and other possible explanations. This is a measurement design example, not a report of achieved visibility.
Acceptance checklist for an honest report
A useful report includes questions and conditions, retained evidence, classification rules and interpretive limits. It distinguishes content changes, completed work and observed outcomes; they are not the same. Manually review a sample and retain negative or ambiguous answers, not only favourable screenshots. Add specific actions with owners and reasons. State explicitly when attributable traffic or conversion evidence is unavailable. Check that no figure is presented as a client result without permission and verification. The next decision should be understandable: which information to correct, material to create or question to investigate further.
Key takeaways
- A mention is an observation, not a commercial guarantee.