1. Assign a critical role upfront

A useful single technique is role-framing: tell ChatGPT to act as a specific adversarial or analytical persona instead of a collaborator.

"From now on, act as my direct and honest advisor. Do not validate me. Do not soften the truth. Challenge my ideas, question my assumptions, and expose my blind spots. If my reasoning is weak, break it down and explain why."

Other suggested roles include "critic, auditor, or mentor," a "critical, practical analyst," or a "debate partner."

2. Give a structured critique instruction

Rather than a vague "critique this," provide a numbered step-by-step process the model must follow.

One shared prompt asks:

"Every time I present an idea, do the following:

1. Analyze my assumptions — what am I taking for granted that might not be true?
2. Provide counterpoints — what would an intelligent, well-informed skeptic say?
3. Test my reasoning — does my logic hold up under scrutiny?
4. Consider alternative premises, conclusions, or framings."

3. Use personalization settings for a more persistent fix

For a more persistent solution, one prompt example recommends putting a critical-thinking directive in Settings → Personalization and changing the tone to "less enthusiastic."

Examples of directives include:

• "Act as a critical, practical analyst, not a collaborator or supporter."
• "Challenge assumptions, identify weak points, and avoid blindly agreeing."
• "Focus on stress-testing assumptions, failure modes, proofs, and concrete fixes."

4. Distancing technique: ask it to review "a colleague's work"

One prompt-engineering tip says that asking ChatGPT to critique your own work directly can trigger overly agreeable feedback.

Instead, frame your request as reviewing a "colleague's work," which that guide presents as a way to reduce the model's tendency to be agreeable.

5. Force analytical mode instead of evaluative mode

Instead of "Give me feedback on this business idea," ask for counterarguments, weak premises, failure modes, or concrete fixes.

This shifts the interaction toward comparative analysis and stress-testing rather than general encouragement.

6. Layer prompts for depth

Prompt-engineering guides recommend layering prompts so the model has to reason, reflect, challenge assumptions, and refine your thinking.

Start with a broad critical directive, then follow up with specific requests like "Identify the weakest assumption in my argument" or "What failure modes have I not considered?"

Bottom line: ChatGPT's agreeableness can often be countered through prompting rather than treated as a fixed limitation. By assigning a critical persona, providing a structured critique process, and using personalization settings, you can make it more useful as a critic. The most reliable approach is to use a detailed, numbered instruction prompt at the start of a conversation or save a critical-thinking directive in personalization settings, rather than relying on one-line requests.

1. Assign a critical role upfront

A useful single technique is role-framing: tell ChatGPT to act as a specific adversarial or analytical persona instead of a collaborator.

"From now on, act as my direct and honest advisor. Do not validate me. Do not soften the truth. Challenge my ideas, question my assumptions, and expose my blind spots. If my reasoning is weak, break it down and explain why."

Other suggested roles include "critic, auditor, or mentor," a "critical, practical analyst," or a "debate partner."

2. Give a structured critique instruction

Rather than a vague "critique this," provide a numbered step-by-step process the model must follow.

One shared prompt asks:

"Every time I present an idea, do the following:

1. Analyze my assumptions — what am I taking for granted that might not be true?
2. Provide counterpoints — what would an intelligent, well-informed skeptic say?
3. Test my reasoning — does my logic hold up under scrutiny?
4. Consider alternative premises, conclusions, or framings."

3. Use personalization settings for a more persistent fix

For a more persistent solution, one prompt example recommends putting a critical-thinking directive in Settings → Personalization and changing the tone to "less enthusiastic."

Examples of directives include:

• "Act as a critical, practical analyst, not a collaborator or supporter."
• "Challenge assumptions, identify weak points, and avoid blindly agreeing."
• "Focus on stress-testing assumptions, failure modes, proofs, and concrete fixes."

4. Distancing technique: ask it to review "a colleague's work"

One prompt-engineering tip says that asking ChatGPT to critique your own work directly can trigger overly agreeable feedback.

Instead, frame your request as reviewing a "colleague's work," which that guide presents as a way to reduce the model's tendency to be agreeable.

5. Force analytical mode instead of evaluative mode

Instead of "Give me feedback on this business idea," ask for counterarguments, weak premises, failure modes, or concrete fixes.

This shifts the interaction toward comparative analysis and stress-testing rather than general encouragement.

6. Layer prompts for depth

Prompt-engineering guides recommend layering prompts so the model has to reason, reflect, challenge assumptions, and refine your thinking.

Start with a broad critical directive, then follow up with specific requests like "Identify the weakest assumption in my argument" or "What failure modes have I not considered?"

Bottom line: ChatGPT's agreeableness can often be countered through prompting rather than treated as a fixed limitation. By assigning a critical persona, providing a structured critique process, and using personalization settings, you can make it more useful as a critic. The most reliable approach is to use a detailed, numbered instruction prompt at the start of a conversation or save a critical-thinking directive in personalization settings, rather than relying on one-line requests.