Prompt You'll Actually Use
Paste this header before you dump notes into AI
Your notes mix facts, guesses, and copied text. Label the pile so AI doesn't flatten them into one voice.
The move Before you paste mixed notes into AI for a summary, add four source labels: FACT, CLAIM, MY TAKE, VERBATIM. Tell it to preserve the labels in the output. Thirty seconds of tagging keeps your guess from sounding like confirmed data.
The short version: AI tends to treat every line you paste as equally reliable. A revenue figure and a hunch come out in the same confident tone. Tag each note by type before you paste, and the summary your team reads will show what's confirmed, what's a claim, and what's your interpretation.
You finished a call, copied two numbers from a report, scribbled your read of the situation, and pasted a quote someone emailed last week. Now you want AI to turn all of it into a summary for your team.
AI tends to treat every line as equally reliable. A revenue figure from the quarterly deck and a guess you jotted between meetings come out in the same confident tone. Your coworkers read the summary and can't tell which is which.
The model can't tell your guess from the data. That's your job, and it takes thirty seconds.
The header
Paste this above your notes before the prompt:
## Source labels
- [FACT] = verified number or confirmed event
- [CLAIM] = something someone said or wrote (not verified by me)
- [MY TAKE] = my opinion or working conclusion
- [VERBATIM] = copied text, do not paraphrase
Summarize these notes for my team. Preserve the source label in your
summary so the reader can see what is verified, what is someone else's
claim, and what is my interpretation. If any note has no label, flag it
as UNLABELED and ask me to classify it before including it.Filled in, it looks like this
## Source labels
- [FACT] = verified number or confirmed event
- [CLAIM] = something someone said or wrote (not verified by me)
- [MY TAKE] = my opinion or working conclusion
- [VERBATIM] = copied text, do not paraphrase
[FACT] Q2 ARR hit $4.1M, up 18% quarter over quarter.
[CLAIM] Jordan said on the call that churn is mostly mid-market.
[MY TAKE] We're probably overweighting enterprise wins because
they're louder in the room.
[VERBATIM] "The renewal pipeline is softer than it looks."
(CFO email, Sep 12).
[FACT] NPS dropped 6 points since the June survey.
Summarize these notes for my team. Preserve the source label in your
summary so the reader can see what is verified, what is someone else's
claim, and what is my interpretation. If any note has no label, flag it
as UNLABELED and ask me to classify it before including it.The labeled summary preserves the distinction
Verified: Q2 ARR reached $4.1M (+18% QoQ). NPS fell 6 points since June.
Attributed claim (Jordan, call): Churn is concentrated in mid-market accounts.
Your interpretation: Enterprise wins may be getting disproportionate weight because they're more visible internally.
Direct quote (CFO, Sep 12): "The renewal pipeline is softer than it looks."
Now your team can see which lines are confirmed data, which are someone's claim, and which are your read of the situation.
What stays human
Deciding the label. AI sorts tagged notes into clear categories. It can't judge whether a line is a verified fact or your working theory.
Try it this week: Next time you collect mixed notes for a summary, paste the four labels at the top first. If you do this weekly, save the header in a Project so it's already there.