Foundations · Lesson 3
Controlling Tone, Length, and Accuracy
You don’t just control what the AI answers — you control how it answers. In this lesson you’ll learn the three controls that stabilise outputs.
Learning objectives
- Control tone (how the AI sounds)
- Control length (how much it produces)
- Improve accuracy (reduce guessing and overconfidence)
- Write one prompt that uses all three controls
The three controls
1) Tone
Tone is not automatic. If you don’t specify it, tone drifts.
- “Use a professional tone.”
- “Be calm and reassuring.”
- “Be direct and no-nonsense.”
2) Length
Length becomes predictable when you set a clear limit.
- “No more than 120 words.”
- “Exactly 5 bullet points.”
- “One short paragraph only.”
3) Accuracy
Accuracy improves when you allow uncertainty and forbid guessing.
- “If you are unsure, say so.”
- “Do not invent facts.”
- “State assumptions explicitly.”
Key insight: reliability increases when you make uncertainty allowed. Forced confidence increases hallucinations.
Practical exercise
Rewrite this vague prompt so it controls tone, length, and accuracy.
Tell me about website security.
Improved version (example)
You are a website security advisor. Explain basic website security to a non-technical beginner.
Use a calm, reassuring tone. Limit the answer to 5 bullet points. Do not speculate; if anything is uncertain, say so clearly.
If you can write prompts like that, your outputs will become noticeably more stable.
Quick quiz
Check your understanding, then move to Lesson 4.
1) What controls tone most directly?
2) What is the best way to control length?
3) Which instruction improves accuracy most?
Score: –