Who asks the question: your buyer or their customer?
Two people can ask the same AI the same kind of question and hold opposite interests. A hairdresser asks how to rank better in AI; their customer asks where to get their hair coloured cheapest nearby. Same mechanics, other side of the table — and in one answer we belong, in the other the hairdresser does.
| The question | Who types it | Who belongs in the answer |
|---|---|---|
| how can I as a hairdresser rank better in AI | the hairdresser — our buyer | ceeme |
| where can I get my hair coloured cheapest nearby | that hairdresser's customer | the hairdresser |
| which tool performs best for AI visibility | our buyer, further along in the decision | ceeme — and this question belongs in our own question mix |
Which audiences are in the measurement?
Ten, and they come from the code — not from a workshop. What separates them is not who they are but what they want to know: one is looking for a solution, another for confirmation, and a third for a flaw.
| Audience | The question they ask |
|---|---|
| The customer | Does this work for my case, and what does it cost? |
| The patient | Can I go there, and how soon? |
| The parent or pupil | Does that programme still exist there, and when do I enrol? |
| The voter | What has that person actually said about this? |
| The sponsor | Is this safe for my brand? |
| The buyer | Who qualifies for this, and can I back that up? |
| The journalist | Is this accurate, and who can confirm it? |
| The investor | What does the market know about this company that I do not? |
| The resident | Which documents do I need, and where do I go? |
| The candidate | Is that a good place to work? |
How much does an answer really differ per audience?
Enough to build decisions on. The same organisation can be the first name for a customer and turn up only as a source for a journalist — and that second case is not a weakness but a different kind of presence. Add both into one percentage and you see neither.
Which audience should I pick if I can only measure one?
The one closest to money or to risk. For an SME that is the buyer, for a hospital the patient, for a school the parent, for an office-holder the journalist. The rest can always be measured later; you start where a wrong answer already costs something today.
Is this not just profiling under another name?
No, and the difference can be named precisely. Profiling is about a person: what do we know about them, what can we show them. This is about a question: how is it phrased when someone asks from that stance. There is no individual, no database and no audience segment anyone falls into.
Frequently asked questions
Can I add an audience myself?
Not as a button, and that is deliberate. An audience decides how a question is phrased, and a badly chosen stance produces a measurement that looks fine and means nothing. If one is genuinely missing, it belongs in the code, and it then appears on this page for everyone.
Do you also measure whether the answer is positive?
Yes, per answer as framing, and broken out per source — the driver tables on the sector pages say so. The difference matters: you know how you are framed, not always by which of the four sources.
Why is this not simply on every sector page?
Because it is a different axis. A sector page is about who buys ceeme; this one is about who reads the answer. A hospital has patients AND journalists AND job applicants, and those three do not fit in one sector's table. Conflating them is also the mistake lib/personas.js explicitly warns about.
Read on
Free, no account and no card.