Question 1

What does Delphi measure?

Accepted Answer

Five sequential gates a buyer passes through when they ask an AI assistant what to buy: Found, Seen, Trusted, Chosen, Bought. The weakest gate caps every gate after it, so it is named as the binding constraint and leads the plan.

Question 2

Which AI assistants does Delphi read?

Accepted Answer

ChatGPT, Gemini, Claude, Perplexity and Grok. Delphi does not read retailer-owned assistants such as Rufus or Sparky and does not imply otherwise.

Question 3

Does Delphi claim the revenue it produces?

Accepted Answer

No. Delphi measures position in AI answers, not revenue caused by it. Projected lift is always stated as a range with the basis it rests on, never as a single figure and never as attribution.

Question 4

How does Delphi handle a number it cannot measure?

Accepted Answer

It says so. A gate that cannot be measured on a run is excluded from the Index rather than counted as zero, and the number of gates scored is published alongside the Index. Every answer is stored verbatim, so any figure can be traced back to what an assistant actually said.

Sign in