How to investigate your business's visibility in AI answers

Start with what a prospective customer is trying to establish before contacting a supplier. They may need to compare approaches, understand a service, check local availability or decide what information to prepare. Use the questions you hear in enquiries and sales conversations to define a small, relevant set.

For an illustrative commercial cleaning business, one question might ask how to compare an office cleaning contract. Another might ask for suppliers in a particular area. These are different tasks and should be recorded separately. A prompt that names your own company can help investigate facts about the business, but it does not show whether an unnamed customer would have been offered that business.

For each observation, keep the complete question, the platform, the date and the context that could affect the answer. Record any stated location, whether the conversation already included information about the business and whether a search or browsing feature was used. Save the returned answer and source links rather than relying on memory.

A useful record lets a colleague understand what happened without recreating the session. It also prevents unlike observations being combined. An answer to a question about suppliers in Manchester is not directly comparable with an answer to a general question about cleaning methods.

A mention is the business appearing by name in the answer. A citation is a source link supporting part of the answer. A visit is someone arriving on the website. These can occur separately: a platform may cite a useful article without recommending its publisher as a supplier, or mention a business without linking to its website.

Decide which observation answers your question. If you want to know whether the company's factual details are correct, inspect the statement and source. If you want to understand referrals, review relevant analytics and completed actions. Do not describe a higher count of mentions as an increase in enquiries unless you have evidence for those enquiries.

If an answer contains an old address or inaccurate service detail, inspect the cited source where one is provided. Compare it with your current website and relevant business listings. Record which statement is wrong, where it appears and what evidence supports the correction.

If the answer does not mention the company, first ask what that observation establishes: the business was absent from that answer in that context. It does not prove the website is blocked, the company lacks authority or one particular schema type is missing. Those are separate questions requiring their own evidence.

Keep the prompts and observation method as consistent as practical. If you change a question, platform or location, note the change rather than treating it as a continuation of the same test. Group observations by the customer task so brand questions, educational questions and supplier comparisons remain understandable.

For a small sample, plain counts are often clearer than a dramatic percentage. “Mentioned in four of the twelve recorded answers in this review” describes the evidence. “We have a 33% share of AI search” would make a much broader claim. The report should identify the sample, date range and limitations.

The free Reeve Consult audit can help investigate the information and technical settings on a sample of your website. It does not submit a prompt set to AI platforms or measure mentions and citations. A clear service page, accurate business facts and appropriate access settings are worth reviewing on their own evidence.

When you commission wider AI visibility research, ask how prompts are selected, which platforms and locations are included, how often answers are observed and how source links are recorded. Agree what the report can establish before using it to justify a content or technical change. That leaves you with evidence you can discuss, rather than a score whose meaning is unclear.