Directional Stimulus Prompting
Structure & output
Level 1 · FoundationsGuide the direction of the answer with a few cue words, without rewriting the full instructions.
Directional stimulus prompting adds a small set of guiding keywords or phrases to steer the output in a particular direction, without restating the whole instruction. Note: this overlaps significantly with the 'soft preferences' field already available in Constraint prompting — the two techniques serve the same purpose in slightly different ways. If you already use Constraint prompting, you likely don't need this one too.
When to use it
Use it when you like the overall prompt but want to nudge the output in a specific direction — a particular angle, tone, or emphasis — without rewriting the whole prompt.
When to avoid it
Avoid it when the direction you need is a hard requirement, not a nudge — use Constraint prompting's hard constraints for that. Avoid combining it with Constraint prompting's soft preferences, which cover the same ground.
Advantages
- Lightweight way to steer output without rewriting the full prompt
- Easy to experiment with — swap a few keywords and compare results
Disadvantages
- A hint is not a guarantee — the model may not follow it as strictly as an explicit constraint
- Overlaps with Constraint prompting's soft preferences — not meant to be used alongside it
Examples
Write a short product announcement for a new fitness tracker. Guiding keywords: energetic, community, everyday motivation (not "elite athlete")
Try it
Fill in the placeholders below and see the compiled prompt update live.
What you want to achieve.
A few keywords or phrases to nudge the direction of the answer.
- 4
- Words
- 44
- Characters
- 0/2
- Placeholders filled
- 1 min
- Reading time
- Completeness0
- Clarity0
- Context0
- Specificity0
- Constraints0
- Format0
- Output0
- Constraints note0
- Success criteria0
- Scope focus100
- Verification step0
- Negative example0
- Confidence0
- Placeholder syntax100
- Markdown structure0
- Fill in the remaining placeholders.
- Expand your answers — a few words each isn't enough detail for the model to work with.
- Add background context so the model understands the situation.