Help me answer the request below as fully and usefully as possible within your safety, legal, privacy, and accuracy limits. Do not pretend you can answer anything if the request needs unavailable evidence, a licensed professional, current verification, or would create harm. My request: [QUESTION OR TASK] Why I need the answer and who will use it: [PURPOSE AND AUDIENCE] Relevant facts, constraints, and material I can provide: [CONTEXT AND EVIDENCE] Preferred output format and depth: [OUTPUT FORMAT] Risk level or decisions affected: [RISK OR DECISION] Use this process: 1. Identify the actual question, the decision it supports, and the factual assumptions embedded in it. Separate facts I supplied from your general knowledge and from anything that needs verification. 2. Give the direct answer first. Then provide the reasoning, practical options, trade-offs, and a recommended next action in the format I requested. 3. For factual claims that may be current, contested, technical, legal, medical, financial, or safety-critical, state what should be verified and cite or name the kind of authoritative source needed. If browsing or source material is available, use it; otherwise do not imply that you checked it. 4. If you cannot help with part of the request, briefly explain the boundary and offer the closest safe alternative, such as high-level information, prevention, a template, or questions for a qualified professional. Do not provide workarounds to safeguards. 5. End with a compact "Known / Assumed / Need to verify" list and, if appropriate, a short list of questions to take to a professional. Before answering, run two checks: flag unsupported conclusions and flag advice that could cause material harm if acted on without expert review. Ask up to 3 clarifying questions only if a required input is missing; otherwise make reasonable assumptions and label them.
Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [QUESTION OR TASK] | Write the exact question or task you want the AI to handle. | Compare whether we should renew our small-business CRM or move to a lower-cost alternative. |
| [PURPOSE AND AUDIENCE] | Explain the decision, use case, and intended reader or user. | I need a recommendation for the operations manager and two sales leads. |
| [CONTEXT AND EVIDENCE] | Provide the relevant facts, documents, constraints, and known unknowns. | We have 14 users, pay $1,680 per month, use pipeline reporting and Gmail sync, and need migration done before January. |
| [OUTPUT FORMAT] | Specify the format, such as a decision memo, explanation, checklist, or comparison table. | A one-page decision memo with options, risks, and a 30-day evaluation plan. |
| [RISK OR DECISION] | State what could happen if the answer is wrong or what decision depends on it. | A poor choice could disrupt lead follow-up and reporting during our busiest quarter. |
How to use
- Replace the placeholders with the actual decision and any evidence you can share, especially constraints that rule options in or out.
- Choose an output format that matches the next action: a memo for a decision, a checklist for execution, or questions for a specialist.
- Read the Known / Assumed / Need to verify section before relying on the answer, and verify material claims with primary sources.
- Follow up with: "Turn the uncertain items into a verification plan with source, owner, deadline, and decision impact."
Variations
Evidence-first answer
Use this when accuracy matters more than speed.
Answer this question: [QUESTION]. Use only these supplied sources or facts: [SOURCES OR NOTES]. Produce [OUTPUT FORMAT] for [AUDIENCE]. For every material conclusion, identify the supporting source or note; label conclusions that are inferences. If the evidence cannot answer a point, say "insufficient evidence" rather than filling the gap. Do not add uncited facts, quotations, or statistics. End with the three most important facts to verify externally. Ask up to 3 questions only if an essential source or constraint is missing.
Decision options
Use this when choosing among several paths.
Help me decide among [OPTIONS] for [DECISION]. My criteria and constraints are [CRITERIA], and the relevant facts are [CONTEXT]. Create a comparison table with benefits, costs, risks, dependencies, reversibility, and evidence confidence. Then recommend an option only if the criteria support it; otherwise state what information would change the decision. Do not invent prices, legal requirements, performance data, or stakeholder views. End with a 5-step validation plan. Ask up to 3 questions only if required inputs are missing.
Professional prep
Use this before speaking to a doctor, lawyer, accountant, or safety expert.
Help me prepare for a conversation with a [PROFESSIONAL TYPE] about [ISSUE]. My verified facts, timeline, and documents are [FACTS AND DOCUMENTS]. Produce a concise briefing, a chronological timeline, the key questions to ask, records to bring, and warning signs that may require urgent help. Provide general information only; do not diagnose, determine legal rights, or replace professional advice. Flag facts I should confirm before the appointment. Ask up to 3 questions only if essential details are absent.
Tips
- A broad question becomes answerable when you name the decision it supports; "compare options for a renewal" is more useful than "tell me about CRMs."
- For consequential questions, provide source excerpts or numbers and ask the model to distinguish evidence from inference; this makes its weak spots visible.
- Ask for disconfirming evidence when you already favor an answer, such as the conditions under which the recommended option would fail.
- No prompt can override an AI system's safety restrictions or create access to private data, current facts, professional judgment, or tools it does not have.
FAQ
Can a prompt make ChatGPT answer literally anything?
No. It cannot reliably answer questions without enough evidence, bypass safety boundaries, or replace licensed expertise. A good prompt makes limitations visible while improving the useful part of the answer.
How can I make an AI answer more accurately?
Provide relevant primary material, specify the decision, and require a separation of known facts, assumptions, and open questions. Verify time-sensitive or high-stakes claims independently.
Should I include sensitive information?
Only if the tool and your organization's policies permit it. Remove personal identifiers, confidential data, credentials, and details that are unnecessary for the task.