Does the AI answer actually name your business?

Before using an AI visibility score to choose software or report progress, open a sample of the underlying answers and confirm that each counted name refers to the intended business. Check alternative spellings, similar company names and links that use your website as a source. Then record whether the answer merely mentions the business or offers it as a suitable choice.

This is a useful trial task for an agency with several clients, a company whose name is an everyday word, or a team reporting on several product brands. A matching rule can be consistent while still counting the wrong entity. A small, documented review gives you concrete questions to take to the vendor.

Write down which business the score should represent

Start with the public trading name, domain and a one-sentence description of the business. Add established abbreviations and previous names only where you intend them to count. List similarly named companies that should stay outside the report. For a local business, record its service category and market as supporting context.

Decide separately how to handle products, subsidiaries and franchise locations. If an answer recommends a parent company, that does not necessarily recommend every product it owns. Combining them can be useful for a group report, but the rule should be explicit and applied to competitors in the same way.

Peec's documentation describes aliases and regular expressions for alternative spellings and more specific matching. Those are examples of controls to ask about, not a claim that every tracker offers them. Ask the provider to demonstrate its own method using your actual naming cases.

Check the sentence, business context and linked destination

Consider a fictional bicycle workshop called Willow Cycles and an unrelated florist called Willow Flowers. The word “Willow” alone cannot establish which one an answer means. “Willow can service a bicycle” provides useful category context; a link to the workshop's domain adds stronger identification. “Willow offers same-day bouquets” refers to the florist.

Read enough of the answer to resolve the reference. A product name, location or domain may clarify an abbreviated name. Where the evidence remains ambiguous, mark it for review rather than forcing a confident match. The following examples are invented; they are not results from a measured tool comparison.

Keep citations and recommendations in separate columns

A domain can support an answer that recommends a competitor. Conversely, an answer can name a business without linking to its website. Peec explains this distinction between source usage and brand visibility. Semrush also documents mentions and domain citations as separate metrics.

In your review sheet, give identity, recommendation and citation separate fields. Keep a brief reason and the original answer reference. A recommendation should reflect what the answer offers for the buyer's question; a positive adjective, page title or reference-list entry is not enough on its own.

Preserve the provider's original classification beside your reviewed decision. That lets you explain disagreements without rewriting the evidence. If the answer is grouped by use case rather than presented as an ordered shortlist, record the group and wording instead of inventing a rank.

Ask what happens when a matching rule changes

Prepare a compact sample containing clear matches, near names, abbreviations, source-only links and uncertain cases. Use the same saved answers to inspect the old and proposed rules where the tool supports that comparison. Reviewing only answers already marked as mentions would miss incorrect exclusions.

Ask whether adding an alias changes historical scores, only future collection, or neither until the provider processes a correction. Keep the date and scope of the change. A rise after broadening the matching rules may reflect reclassification rather than more AI answers recommending the business.

A manually labelled sample can reveal specific disagreements. It does not establish accuracy across all prompts, languages or engines. For repeated collection, use the separate fair AI visibility test to keep question wording, market and engine coverage consistent.

Use the trial to establish an auditable workflow

Ask the vendor to open an underlying answer, explain a disputed match and show where the original citation leads. Confirm who can correct a business identity and how that correction appears in a client report. If a control is unavailable, decide whether a manual review process is workable for your team.

With Reputably's AI visibility tracking, review collected answers and cited sources for the customer questions you monitor. Use a Reputably trial to check your business identity against actual answer text. This method does not imply a built-in alias editor or historical rescoring feature.

Compare software using the AI rank tracking buyer's guide, and document the accepted counting rules in your agency reporting brief. Keep search clicks, landing-page engagement and qualified conversions alongside the visibility report as separate outcomes. Correctly identifying a recommendation does not prove that it produced a visit or sale.