AI Prompt Settings for Meeting Summaries (Length, Format, Language)

Why the Same Summary Prompt Wastes Time

A meeting summary that is perfect for a busy exec is useless for an engineer who needs action items. The fastest way to control AI meeting summaries is not a new tool — it is three settings you put in the prompt: length, format, and language. Get these wrong and you either drown in transcript or miss the decision that mattered. Get them right and the same audio note becomes exactly the artifact each reader needs.

This article is a settings table for meeting summaries. Pick the row that matches your reader, then copy the snippet. If you already capture raw notes and just need to compress them, the approach in meeting notes summary prompt pairs well with these settings, and research summary settings covers the same knobs for longer documents.

Meeting Summary Settings Table

When to Use Each Setting

Exec one-pager. Use when the reader is senior and time-poor. They want the single decision and the three things that follow. No agenda, no “great discussion.” Anything longer gets skipped.

Action list. Use immediately after standups, sprint planning, or client calls where the value is who does what by when. The table format drops straight into a task tool.

Full brief. Use for kickoffs, reviews, or meetings where context matters and people will read days later. The four sections preserve why a decision was made, not just what it was.

Transcript-light. Use when the meeting mixed languages or when you want the model to keep the original wording for nuance. It still compresses, it just does not translate.

Bilingual mirror. Use for cross-border teams where some readers need English and some need the original. Two columns avoid confusion about what was actually said versus what it means.

Prompt Snippets by Setting

Exec one-pager

Summarize this meeting for an executive who did not attend.
Rules:
- Total length about 120 words.
- First line: the single decision made.
- Then exactly 3 bullets: what changes, who is affected, next milestone.
- No agenda recap, no pleasantries.
- Language: clear English, native level.
Transcript: [PASTE]

Action list

Turn this meeting into an action list.
Output a table with columns: Owner, Task, Due date.
Rules:
- Only include concrete tasks with a clear doer.
- If no due date was said, write "TBD".
- Drop pure discussion with no action.
- One row per task, no paragraphs.
Language: plain English. Transcript: [PASTE]

Full brief

Write a meeting brief from this transcript with these sections:
1. Context (why this meeting happened)
2. Discussion (key points, 3 to 5)
3. Decision (what was agreed)
4. Risks / Open questions
Rules:
- 300 to 500 words total.
- Keep the "why" behind each decision.
- Neutral tone, no opinion.
Language: English. Transcript: [PASTE]

Transcript-light

Summarize this meeting but keep the original language of what was said.
Rules:
- Medium length (about 200 words).
- List key points in the order they happened.
- Do not translate; preserve source-language terms where they matter.
- Keep names and numbers exact.
Transcript: [PASTE]

Bilingual mirror

Summarize this meeting as a two-column mirror.
Left column: key points in [SOURCE LANGUAGE], kept faithful.
Right column: English translation of the same points.
Rules:
- Medium length, ~6 to 10 rows.
- Each row is one point, original then translation.
- Do not add commentary.
Transcript: [PASTE]

Common Mistakes With Summary Settings

1. Picking one format for everyone. The exec who wants a decision and the engineer who wants owners-and-due-dates are different readers. Match the row to the reader, not to your habit.

2. Letting the model translate when you needed the original. If nuance lives in the source language, use the Transcript-light setting instead of the default English summary, or you lose the exact wording that matters later.

3. Asking for “a summary” with no length cap. Unbounded summaries grow to half the transcript. Always state a word count or row count so the model compresses instead of paraphrasing.

4. Skipping the action table after a decision meeting. Decisions without owners go nowhere. Even when you use the Exec one-pager, paste it once more into the Action list snippet so the team has a to-do view.

Quick Reference Card

The Trade-Offs

Short summaries save reading time but hide context, so a one-pager alone can confuse someone who was not in the room. Long briefs preserve nuance but get ignored under load. The practical move is to generate two outputs from one transcript: a one-pager for leadership and an action list for the team. Both snippets above run on the same paste, so the cost is one extra prompt, not double the work. For more ways to route AI outputs to the right person, productivity prompt list has templates for follow-ups and task handoffs.

Sources & Notes

Prompts tested with ChatGPT (GPT-4o and GPT-4.1), free and Plus tiers, in October 2026. The length/format/language knobs are prompt-level controls and work in any chat model; no API or plugin required. Slug and category mapping follow the guunn.com AI Prompts plan (Productivity = 960). No live APIs were called during article creation; snippets are copy-paste text for the reader.