How to Write Probe Notes for an AI Moderator
A probe note tells the AI moderator what to do after an answer. These rules show how to write notes that produce specific follow-ups and stop at the right time.
Overview
A probe note is the instruction under a guide question. It tells the moderator when to ask a follow-up, what to ask and when to move on. A person who moderates can see that an answer is thin and improvise. An AI moderator has only the question, the note and the answer.
A vague note gives a vague follow-up, such as "Can you tell me more?", in every interview. A specific note gives a follow-up that targets the gap in this answer. The difference is small in one session and large across a full study.
In User Evaluation, each guide question has a field called "Follow-up if the answer is thin". A question can have up to five follow-ups. The probe depth setting sets the follow-ups on each answer: none (Light), one (Standard) or two (Deep).
Rules
- 01
Write each probe note as a condition and an action.
The moderator must know when a follow-up is necessary, not only what to ask. Without a condition, it asks the follow-up even when the first answer was complete.
Do notProbe on pain points.
DoIf they name a problem but give no example, ask when it last happened. If they already gave an example, move on.
- 02
Name the exact detail that is missing.
"Dig deeper" does not tell the moderator which direction is deeper. Name the fact that your analysis needs: a time, a person, a step, a workaround or a cost.
Do notDig deeper into their answer.
DoAsk what they did right after the error appeared, and who they told about it.
- 03
Give every note a stop rule.
An AI moderator does not get tired or embarrassed, so it can ask again and again. A stop rule prevents repeated follow-ups that annoy the participant and add no data.
Do notKeep asking until you understand why they switched.
DoAsk one follow-up only. If they still say "I don't know", accept it and go to the next question.
- 04
Turn general labels into a moment.
Words such as "confusing", "slow" or "clunky" are labels. You cannot design a fix from a label. The note must tell the moderator to ask for the moment behind the label.
Do notIf they say it is confusing, ask why.
DoIf they say "confusing", ask which screen they were on and what they expected to happen there.
- 05
Keep your hypothesis out of the note.
The moderator often copies words from the note into its follow-up. If the note names the answer you expect, the participant hears that answer as a suggestion.
Do notAsk if the new dashboard made the weekly report easier.
DoAsk how they made the weekly report last time. Do not name the dashboard unless they name it first.
- 06
Explain internal names and terms in the note.
The moderator knows only what the guide tells it. It does not know your team's project names or what participants call a feature.
Do notAsk about the V2 flow.
DoThe "V2 flow" is the new three-step sign-up. Participants call it "the new sign-up". Use their word, not ours.
- 07
Keep each note to one or two conditions.
A note with many branches is hard to follow. The moderator can stack several follow-ups into one turn, and the participant answers only the last part.
Do notAsk about price, setup time, who approved it, which tools they compared, how long the trial was and whether support helped.
DoIf they do not say who approved the purchase, ask who had the final say. Do not ask about price here.
- 08
Read the follow-ups that the moderator asked, not only the notes you wrote.
A note can look clear and still produce a strange follow-up. The transcript of a pilot shows the real words that the moderator used.
Checklist before launch
- Each probe note starts with a condition, such as "If they…".
- Each probe note names one missing detail: a time, a person, a step or a workaround.
- Each probe note has a stop rule, or says "Do not ask a follow-up".
- No note contains the answer you hope to hear.
- Internal names and product terms are explained in the note.
- Context questions have the note "Do not ask a follow-up".
- The probe depth setting matches the number of follow-ups that the notes expect.
- You ran a pilot and read each follow-up the moderator asked.
Questions
A probe note is an instruction under a question. It tells the moderator when to ask a follow-up, what to ask and when to stop.
Yes. An AI moderator can write its own follow-ups from the answer. But these follow-ups are often general. A probe note makes the follow-up target the detail that your analysis needs.
No. Write it as an instruction with a condition. The moderator then puts the follow-up into natural words that fit the answer.
One or two sentences. If a note needs more, the question probably asks about more than one idea. Split the question.
Ask real people your question
The AI moderator interviews screened people by voice or text. Each claim in the study report links to its clip. The first 3 interviews are free.