Competitive Analysis

Competitive analysis, in the context of AI visibility, is the record of who else an assistant names when your customers’ questions are asked, how often, in what position and with what reasons. It turns a single score into a picture of the field the score was earned in.

What the analysis records

Every answer in a question panel contains more than a yes or no about one business. It contains the names that were given instead, their order, and usually a stated reason: a review count, a location, a specialism, a price. Recording those for every answer produces a table of who is recommended alongside whom, per question and per assistant. On this site that is part of the standard measurement described on the methodology page, asked of one assistant, Claude: for each answer, whether the business was named, whether it came first, and who else was named.

One thing that record shows, and a results page does not, is what kind of thing is being named. In this site’s probe of 9,240 businesses, 99.4% of the names Claude returned were businesses and 0.6% were directories or aggregators, as reported in the 9,483-site audit. The competitors in an AI answer are other businesses, not the listing sites that fill the first page of search.

Reading the result

The useful question is not who ranks above you but what the named businesses have on their sites that yours lacks. The names come with reasons, and the reasons point at pages: an assistant that says a competitor is well reviewed read that somewhere. The audit found the businesses that were named were, on inspection, the ones whose sites had answer-shaped content and a consistent name, address and phone number. A competitive analysis that stops at the names list has skipped the part that can be acted on.

A common question

Are directory sites the competition in an AI answer, or other businesses?

Other businesses, by a wide margin. In the probe of 9,240 businesses, 99.4% of the names Claude returned were businesses and 0.6% were directories or aggregators, close to the reverse of a search results page, where listing sites fill the first screen. The practical effect is that the record of who gets named alongside you is a list of trading competitors, and the reasons attached to their names point at pages on their sites rather than at a profile someone else maintains.

Sources

Probe and audit figures are this site’s own, computed from its audit records; one assistant, one question per business, sample drawn per metro and trade. Mechanism: Aggarwal et al., “GEO: Generative Engine Optimization” (KDD 2024).