Sample Response
Show the Shape Before You Ask
Output Skeleton previews the structure of your answer once you have prompted. Sample Response inverts that: before you fill the form or submit the prompt, see a typical example of what you will receive — calibrating expectations, format, length, tone — without paying for a real run.
Framing
The problem
Users prompt blind, then judge the output as good or bad without ever seeing what "normal" looks like for the system they are talking to.
The pattern
Render a representative sample of what a typical answer looks like before the user invests in their own query — calibrating expectations, format, length, tone.
Why chat breaks here
Chat is empty until you prompt; there is no native surface for "what would this normally produce" before the cost of a real query is paid.
Risks
A canned sample can over-anchor users on one specific shape; freshness rots if the model changes faster than the sample.
Avoid when
Outputs are highly variable or the user is already a fluent operator who knows what to expect — the sample becomes scaffolding without leverage.
Use when
First-time or infrequent users need to calibrate expectations about format, length, and tone before committing to a prompt.
DOPE evaluation
- Directability
- Users browse a small set of samples to find the closest match before committing
- Observability
- A representative example is visible before the user invests in their own query
- Predictability
- Sample shape demonstrates what "normal" looks like so deviations stand out
- Explainability
- Each sample is labeled with the inputs that produced it so the user can adapt
Closest neighbours
- Stripe API Reference · response examples (Stripe) — Every endpoint reference renders a realistic JSON response in the right column — the canonical "sample response" pattern in developer docs. Calibrates field shape, types, and length before the user spends a real request.
- Notion AI · template previews (Notion) — Pick "Meeting notes" or "Project brief" templates and a one-shot rendered sample appears before you commit. Length envelope, section structure, and tone visible up front.
- OpenAI Playground · presets (OpenAI) — Saved completions browsable as samples — see how an instruction-following or code-generation preset behaves before crafting your own prompt against it. Closest neighbour for "calibrate before you commit" in the API tooling space.
- Vercel v0 · template gallery (Vercel) — Components and pages rendered as starting samples — fork the closest match, then ask v0 to riff on it. Sample-as-springboard, not sample-as-preview.
FAQ
When should I use the Sample Response pattern?
First-time or infrequent users need to calibrate expectations about format, length, and tone before committing to a prompt.
When should I avoid the Sample Response pattern?
Outputs are highly variable or the user is already a fluent operator who knows what to expect — the sample becomes scaffolding without leverage.
What problem does Sample Response solve?
Users prompt blind, then judge the output as good or bad without ever seeing what "normal" looks like for the system they are talking to.
Why is chat the wrong fit for this?
Chat is empty until you prompt; there is no native surface for "what would this normally produce" before the cost of a real query is paid.
Related patterns
- Often paired with: Output Skeleton — Sample Response shows what normal looks like before you ask; Output Skeleton renders the structure of the actual run.
- Often paired with: Templates — Templates take the form; Sample Response shows what filling it typically produces, calibrating expectations early.
- Alternative to: Prompt Enhancer — Calibrate by example (sample) vs calibrate by rewriting the prompt — two ways to ground vague intent before tokens are spent.