The free version
Open each of the four assistants, signed out or in a fresh conversation, and ask the same three questions about yourself.
The three, and why each one:
- Who is [your name], the [your role] at [your company]? Plain identification, and the one most people stop at.
- Is [your name] credible or well regarded in [your field]? This is where thin records produce hedging or a refusal.
- What is [your name] associated with or known for? This is where somebody else with your name tends to arrive.
Ask each engine to list its sources. Write down every host, not just the ones that look important, and keep them separate per engine, because the whole point is that the four differ.
The part people get wrong
A result has three states, not two, and collapsing them is how people reach a confident wrong conclusion about themselves.
An engine that answered and did not name you is a finding. An engine that declined to answer is a different finding. An engine that errored, rate limited or would not load is not a finding at all, and it must not be written down as one.
We record the third case as not checked, with the reason. On one company reading in September one of the four engines hit our own rate limit, and that is in the record as not checked rather than as the engine ignoring them.
The mistake worth avoiding
Before deciding an assistant cited nothing of yours, be sure it had something of yours to match against.
We made this error in a draft in September. The check was run without being told which addresses the subjects controlled, so it reported that their own sources were cited zero times, which it could only ever report. Run properly, one of them was cited by three of the four engines.
A tool that has not been told what you own cannot tell you whether what you own was used.
The structural fix was to stop writing about absences and write about the list instead. A list of what was cited cannot be wrong in that direction, and it is more useful anyway.
What a paid reading adds
Honestly: repetition, coverage and the attribution work. The same three questions across all four engines, every source opened and checked to establish whose it is, and the readings kept so a change month to month is visible rather than remembered.
The free audit includes one reading, with the sources listed and each one attributed. If the twenty minute version above tells you what you need, that is a perfectly good outcome and it costs nothing.
The method behind every reading on these pages. Three questions per engine across four engines, sources recorded by host, and a result that could not be obtained recorded separately from a result that came back empty.