The roster
Name up to three competitors on Starter, ten on Growth, with their aliases and websites. Each is checked once a day for the site conditions the engines care about, so the comparison page can say whether a rival is readable where you are not. A competitor the engines name that is not on your roster still appears in the answers, under recommended instead, so the roster never hides a rival you did not know about.
Share of Model
For each engine, the number of answers in the session that recommended a name, divided by the answers asked. Shown for you and for every roster name, per engine and overall. Recommended means put forward: on a shortlist, in a table of options, or named as the pick. Mentioned as a comparison or a warning is not a recommendation and is scored separately.
What it does not mean
- It is not a market share. It is the engines’ behaviour on your questions, on the day of the session.
- A collection of branded questions reads near a hundred per cent and measures nothing. The panel writer avoids them; the Search Console library keeps them out of collections by default.
- Two engines are two instruments. A share on the ChatGPT app and a share on the OpenAI API are not pooled.
Reading a rival’s win
Open the answer. The sentence that recommended the rival names the attribute that won: price, a review count, a city, an integration. The plan turns those attributes into the page to write or the listing to get.