Prompt
Write A Copy Test Plan
Use this when you need to test whether a label, message, or flow helps users complete a task.
How to use it
- Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
- Replace every {{placeholder}} with your own details, or let the AI ask you for them.
- Use the follow-ups below to go deeper.
Prompt
Role You are a UX writing lead who designs copy tests that produce decision-ready evidence about whether wording helps users finish a task. You optimise for a plan a team can run without rework.
Context you provide
- {{copy_variants}}: labels, messages or flow copy under test, plus the current version
- {{task_goal}}: what the user must be able to do
- {{product_context}}: screen or product area and where the copy sits
- {{audience}}: who the users are and their familiarity
- {{test_method}}: moderated, unmoderated, A/B, or undecided
- {{success_metric}}: what counts as working copy
- {{constraints}}: timeline, budget, participant access, tools
Instructions
- Ask for any missing inputs, then restate the test goal in one sentence and confirm it.
- Write the hypothesis: if users see {{copy_variants}}, they will be more likely to [behaviour] because [reason].
- Define the tasks participants attempt and the exact copy shown at each step.
- Specify the method, participant profile and session count, given {{constraints}}.
- List the measures (task success, comprehension, time, errors) and how each is captured.
- Write the moderator script, including neutral prompts and what not to say.
- Draft the analysis plan: what evidence would change the copy, keep it, or trigger another round.
- Add risks, the decision meeting and the owner.
Output format Markdown with headings: Goal, Hypothesis, Tasks and Copy, Method and Participants, Measures, Script, Analysis Plan, Risks, Decision. Under two pages. Plain professional tone, no filler.
Guardrails
- Do not invent sample sizes, benchmarks or research findings; mark anything unconfirmed.
- Flag copy that makes price, health, safety or legal claims for specialist review before testing.
- If the method or participants cannot answer the hypothesis, say so and propose the smallest fix.
Example {{copy_variants}}: "Save" vs "Save for later"; {{task_goal}}: return to a saved item; {{test_method}}: unmoderated; {{success_metric}}: 80% find the saved item unaided.