Ask ChatGPT to identify your assumptions, test them against the available evidence, and give the strongest reasonable counterargument—not simply to “disagree.” For a more direct, constructive tone, try the Candid personality; for a preference you want applied across chats, use Custom Instructions. Neither setting guarantees accuracy, so treat the response as a way to examine an idea, not as proof that it is right or unbiased.
Use a prompt that asks for evidence-weighted criticism
Be explicit about what you want ChatGPT to examine and how you want it to respond. A useful instruction asks it to separate facts from inferences, identify uncertainty, and explain what evidence might change its assessment. This is suggested prompt wording, not an official OpenAI recipe or a guarantee of behavior:
When I share an idea, plan, or conclusion, help me examine it rather than automatically affirming it. Identify the assumptions it depends on; distinguish facts from inferences; give the strongest reasonable counterargument; point out missing information and uncertainty; and suggest what evidence would change the assessment. Keep the critique constructive and proportionate. If the evidence supports my view, say so.
This works better than asking ChatGPT to “be brutally honest” or “always disagree”: those instructions describe a tone or outcome, but not the analysis you need. OpenAI’s prompting guidance recommends clear, specific requests with enough context, explicit tone guidance, and refinement when the answer misses the mark.
#1 Best Overall
Choose where to set the preference
Use a one-off prompt when you want a particular idea or decision examined. Put a stable preference in Custom Instructions if you want it to apply across chats. The Candid personality is an optional tone setting for direct, constructive responses.
| Option | Best for | Where to set it |
|---|---|---|
| Prompt in the conversation | One idea, plan, or decision; include the relevant context and evidence with the request. | Type the review instructions in the current chat. |
| Custom Instructions | A recurring preference for how ChatGPT should review your ideas. | Web/Desktop: Settings > Personalization, enable customization, then enter instructions in the Custom Instructions field. iOS/Android: Settings > Customize ChatGPT, enable customization, then enter instructions. |
| Candid personality | A direct and encouraging style that can call out risks, gaps, and trade-offs while offering next steps. | Web: profile menu > Personalization > dropdown beside Base style and tone. iOS/Android: profile menu/Settings > Personalization > Base style and tone. |
OpenAI describes Custom Instructions as a way to tell ChatGPT what it should consider and how you prefer it to respond; they apply immediately across chats. OpenAI Academy recommends using them for stable preferences such as tone and guardrails, while putting the specific task and context in the prompt. See OpenAI’s Custom Instructions guide and OpenAI Academy’s guide to personalizing ChatGPT.
Use Candid for tone, not as a substitute for a clear request
OpenAI describes Candid as “Direct and encouraging. Gives honest feedback with clear next steps.” It may suit a review where you want risks and trade-offs stated plainly. A personality changes style and tone; it does not change ChatGPT’s capabilities or safety rules, and it does not control every kind of generated content. Your prompt still needs to say what to analyze. Read OpenAI’s personality guide for the current setting details.
Keep persistent instructions within the character limit
OpenAI’s Help Center lists a 1,500-character limit for Free and Go users and 5,000 characters for Plus, Pro, Enterprise, Business, and Education users. These product limits can change; check the current Custom Instructions guide if the exact limit matters. Keep the persistent instruction focused on your general preference, and provide the specifics of each decision in its chat.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Ask for a structured review of a real decision
For a consequential question, give ChatGPT the relevant facts, your goal, and any constraints rather than relying on a general preference alone. Then request a review in parts:
- State the decision or conclusion. Explain what you are considering and what outcome matters to you.
- Provide the evidence and constraints. Include relevant facts, assumptions, deadlines, and what you do not yet know.
- Ask it to identify assumptions. For example: “What am I treating as true without establishing it?”
- Request the strongest counterargument. Ask for the most reasonable case against your view, not a list of objections for its own sake.
- Separate support from inference. Ask which claims follow from information in the chat and which depend on assumptions or missing facts.
- Ask what could change the assessment. This helps identify useful evidence to gather instead of treating the first answer as final.
Refine the answer if it is too agreeable or too contrarian
Read the response for whether it actually weighs the evidence. If it mostly echoes your view, ask for a stronger counterargument and the assumptions that would make your conclusion fail. If it objects to everything regardless of support, ask it to compare the evidence on both sides and say when your view is better supported. OpenAI recommends inspecting responses and refining prompts; its Prompting guidance also advises stating the task, adding useful context, describing the desired output, and naming priorities.
ChatGPT’s response Characteristics can adjust warmth, enthusiasm, formatting, and emoji use. Those controls work alongside personality, Custom Instructions, and saved memories, but they tune style rather than add reasoning capability. See OpenAI’s guide to personality and response characteristics.
Use critical feedback as a review aid, not a verdict
A more direct tone or a request for counterarguments does not establish that an analysis is correct or unbiased. OpenAI notes that personality affects style, not capabilities or safety rules, and instructions and context can influence how the selected personality comes through. For important factual claims, check reliable sources independently and give ChatGPT the evidence you want it to assess. A useful answer should make its reasoning and uncertainty easier to inspect—not remove the need for your judgment.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
