Copperline Credit Union 21-Scan report

We count three different things, and only the first one sends you a member. Recommended means the written answer puts you forward as a choice. Named in the answer also covers a question that already contained your name, and a head-to-head that names you while telling the reader to pick someone else. Named anywhere additionally counts the map-style list beside the answer, which comes from business listings. Appearing there means you have a branch nearby, nothing more. The headline is the first number.

These assistants are not deterministic. Ask twice and you can get two different lists. So where we have asked a question more than once, we report how many of those runs named you rather than a bare yes or no: an answer that only lands sometimes is not a position you hold. Where an engine has been run only once, we say so below, and we cannot yet tell a stable answer from a lucky one on that engine.

This scan: Perplexity 2 runs per question · Google 1 run per question · ChatGPT 2 runs per question · Claude 2 runs per question.

The headline blends the engines we have measured: a question counts once if any measured engine recommended you. We never average engines into an invented score, because they disagree, and the disagreement is information. The engine breakdown above shows each one on its own. Two things are always left out of the math: engines we have not run yet, and questions where an engine showed a map instead of an AI answer. Neither is a miss.