Why the one-click version does not work
Ask an assistant the same commercial question ten times and the brands named will change, the sources cited will shift and the ordering will move. Any tool returning a single result is showing you one draw from that distribution and presenting it as a position.
There is a second problem that affects you personally rather than the tool. Personalisation and conversation memory bias answers toward whatever the account has recently looked at — frequently the very site being checked. Testing your own brand while signed in usually measures your browsing history rather than the market, and it always flatters.
Running it properly means paid API calls across several engines, repeated over a hundred or more prompts, on a schedule. That is a funded service rather than a free browser tool, which is why this page explains the method instead of pretending to execute it.
The method you can run yourself
- Build a panel of 100–200 commercial prompts from how buyers actually speak — sales call recordings, support tickets, customer interviews. Not keyword lists reworded as questions; the phrasing differs substantially and it changes the results.
- Fix a competitor set at the same time and do not change it later. Share of voice against a moving comparison set is not a measurement.
- Run cold. Fresh session, no signed-in account carrying memory, no prior turns, consistent geography. This is the step most often skipped and the one that most affects validity.
- Write the scoring rules before the first run. Score three things separately: named at all, cited with a link, and whether the mention is favourable, neutral or a caveat. They move independently and a blended score hides that.
- Repeat weekly across the engines your buyers actually use. Daily generates noise that invites over-reading; monthly leaves too long to attribute a change to anything.
- Report direction and proportion, not attribution to a single action. Too many variables move at once for that to be honest.
If you would rather we ran it
That is the GEO baseline audit: a fixed panel, run cold across five engines on a weekly schedule, scored by written rules, with the methodology handed over so you can check the working or run it yourself afterwards.