Design a claim-and-evidence structure that moves a specific reader toward a defined outcome
5 min setupTested with:ChatGPTReviewed: 2026-08-28
1
Add your context
Your text stays in this browser. AILesson Prompts does not send it to a model or server.
2
Your prompt
Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt
Create an article outline for the stated goal and audience.
Desired change in understanding, decision, or action:
[goal]
Readers, prior knowledge, questions, stakes, objections, and reading context:
[audience]
Verified material, required points, exclusions, length, format, tone, and deadline:
[material]
Use only supplied facts as claims. Start with the reader's need and design a logical progression toward the article goal. Return: one-sentence promise; scope and non-goals; working thesis; 2–3 structural options with trade-offs; recommended outline with each section's reader question, purpose, key claim, supporting evidence, example or visual need, transition, and approximate share of length; objection and uncertainty placement; introduction and conclusion jobs; evidence gaps; and a drafting order. Avoid generic headings, repetitive sections, clickbait, and claims whose evidence is absent.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.
From input to outcome
A worked example
See how concrete context turns this recipe into a usable result
Actual input
Article goal
Help managers of 20–100-person organizations decide whether to introduce a written meeting-decision log and start a two-week trial.
Audience and context
Busy non-specialist managers. They know meetings lose decisions but worry another template creates administration. They will scan a 1,200-word article on a professional blog. They need a low-risk next step, not software advice.
Material and constraints
Verified internal sample: in 36 project meetings, 11 later disputes involved unclear decisions or owners. After one team trialed a log for six weeks, 19 of 21 decisions had owner/date fields; no causal productivity result was measured. Required: define a decision log, show minimum fields, address overhead, explain when not to log, give a two-week trial. Exclude claims that logs improve productivity. Calm practical tone.
Example output
Promise: show managers a minimal, reversible way to make consequential meeting decisions traceable without turning every conversation into paperwork. Thesis: a decision log is useful when it records only consequential decisions, evidence, ownership, and change—not as a full meeting archive.
Alternative structures: problem→tool→trial is fastest for scanning; story-led is engaging but risks overgeneralizing one team; field-by-field guide is practical but may delay the decision. Recommend problem→tool→trial.
“What keeps getting lost?” 15%: define the 11/36 dispute observation and its limits; transition from memory failure to a narrow artifact. 2. “What is a decision log?” 20%: minimum fields—decision, date, decision-maker, rationale/source, owner/due date if action, status; visual: one anonymized row. 3. “Will this create more admin?” 20%: distinguish decisions from discussion; state what not to log; use 19/21 completeness only as trial evidence, not productivity proof. 4. “When should we skip it?” 10%: routine reversible choices and sensitive content governed elsewhere. 5. “How do we test it safely?” 25%: two-week scope, owner, review questions, stop/adjust rules. Conclusion 10%: invitation to run the reversible trial.
Evidence gaps: time spent maintaining the log and dispute trend after adoption. Draft sections 2 and 5 first, then evidence, objection, intro, and conclusion.
Why this works
1
Reader questions give each section a necessary communication job instead of merely naming a topic.
2
Evidence fields expose unsupported sections before prose makes them sound finished.
Check the result
Does every section answer a distinct reader question and advance the article goal?
Are claims, evidence, objections, and uncertainty placed explicitly?
Do estimated section shares fit the stated length and avoid repetition?
ChatGPT