TL;DR - Most weak output is a format problem: too long, wrong shape, wrong voice. Direct the format, tone, and length explicitly and quality jumps immediately.
Why it matters
Beginners accept whatever shape the model returns. Pros specify it. The fastest path from "meh" to "professional" is telling the model exactly how the answer should look and sound.
Worked example - one input, three shapes
From the same notes:
"...turn this into an executive email." -> 5 tight sentences, formal.
"...turn this into a bulleted summary." -> 5 scannable bullets.
"...turn this into a Slack message." -> 2 lines, casual, an emoji.Same content, three deliverables, just by changing the format line.
Steal this - the control trio
Format: bullets | table | numbered steps | JSON | email
Tone: formal | friendly | confident | cautious | playful
Length: one sentence | under 100 words | exactly 5 bulletsCommon mistakes (and the fix)
- No length limit -> walls of text. Fix: "under 100 words" or "exactly 3 bullets".
- Wrong format for the channel. Fix: match it - a table for comparisons, bullets for scanning.
- Tone mismatch. Fix: name the tone, or paste a sample and say "match this voice".
Good to know
Asking for a table or JSON is especially powerful: tables make comparisons instantly usable, and JSON is what lets AI feed spreadsheets and automations (Level 4-5). All major assistants honor explicit format requests - and for data, ChatGPT and Gemini can output straight into Sheets.