A recommendation is the end of a path. Every step before it can be measured.
An assistant has to know you, understand you, find you relevant, back it with sources, and shortlist you. We measure every step, tell you whether the question is worth fighting for, and which words to change to move it. Each measure returns its own verdict, so a flat result always points at something specific.
We put your buyers’ questions to ChatGPT, Gemini and Claude, several times each, and record what comes back: who is named, in what order, from which sources, and how each assistant describes and files your business. The same questions run again after you make changes, so a second reading can be compared to the first.
Assistants answer differently on identical prompts, and none of them publishes how a choice was made. We measure across runs and models and report what recurs. We never promise selection, and every reading is dated.
Choose a plan after your first reading.