Nine questions to ask an AI-visibility vendor before you resell it

White-labelling a measurement product means your client's trust in the number is really trust in you. Here is the due-diligence list we would want an agency to run on us.

The nine

1. How was this number acquired? Grounded API, consumer-interface scraping, or a SERP vendor — and does the answer change per engine? All three are legitimate. Not being told is not.
2. Can I see the answers behind this figure? Count the clicks. If the raw answers are not reachable from the metric, the metric is unverifiable.
3. How many runs is this figure based on? If the interface cannot tell you, it is showing you a number it cannot support.
4. Is this a range or a point? A point estimate off a small sample is the category's default failure mode.
5. Did a web search actually fire on these runs? Grounded APIs skip search a large fraction of the time. A tool that does not record this cannot distinguish “the engine looked and you weren't there” from “the engine never looked”.
6. How is location handled? Injected into the prompt text, or passed as a request parameter? The first is not localisation.
7. What happens when an engine fails mid-cycle? Partial coverage reported honestly, or a quietly narrower aggregate presented as a full one?
8. Who chooses the competitor set, and is it visible to the client? Share of voice without a published denominator is not a measurement.
9. What is included, and what is a per-engine add-on? Engine add-on pricing is the category's standard margin trap: the headline price covers two engines and the coverage you actually promised the client costs three times as much.

Two more that are about you, not them

• Whose brand does the client see? Logo, colour and domain, or a co-branded dashboard that teaches your client the vendor's name.
• What happens to the data when the contract ends? Both directions: can you export it, and can you have it erased.