How I collect AI answers

A teardown is only worth reading if you can trust the answers in it. This page explains how I get them, so you can judge for yourself.

What I look at

I check four places where buyers get answers: ChatGPT, Gemini, Perplexity and Google. For Google, I look at the AI answers (AI Overview and AI Mode). I also note who holds the top three regular results for “you vs competitor”.

I can add Claude, Grok or Copilot on request. That makes sense if you know your buyers use one of them.

How I set up each check

The questions I ask

For each competitor, I ask six questions. I type them exactly the same way in every engine.

  1. One plain comparison. “You vs competitor”.
  2. Three buyer-style comparisons. I pick three of these: for a certain type of company, for a certain use, on the thing the competitor wins on, or whether it is worth switching.
  3. Two broad questions. “Best tool for this type of company”, and “alternatives to the competitor”.

Six questions in four engines make 24 answers to read. I choose the buyer wording from your own website: who your customers are, what your case studies say, how you talk about pricing. The teardown tells you what I assumed. If I got it wrong, I rerun it.

The three numbers I report

  1. Who the answer favors. Across the 16 answers to the comparison questions, how many favor you, how many favor the competitor, and how many stay neutral.
  2. How often you are named. Across the 8 answers to the broad questions, how many name you and how many name the competitor.
  3. Where the sources come from. Of all the sources the answers cite, what share comes from your site, from the competitor’s site, and from other sites such as list articles, review sites and forums.

“Favors” is a judgment call. I make a first pass, then I read each answer and decide myself.

What this cannot tell you

Want to see it on your company?

The free teardown uses this method on you and one competitor. Ask for one.