The useful result
Get a useful critique of a plan or draft from three differently framed prompts.
- Get a useful critique of a plan or draft from three differently framed prompts.
You share a business idea or a draft and the chatbot replies that it is a great plan. Share the opposite and it agrees with that too. Assistants can lean toward agreeing with the person in front of them, and a hint about what you want is often enough to tilt the answer.
You cannot switch this off, but you can design around it. Hide your preference, ask for the case against, and compare a second opinion that never saw your enthusiasm. Anthropic suggests running the same prompt several times and comparing; the same idea works for opinions.
The examples below are illustrative exercises, not recorded results from any tool. Sources were checked on 24 October 2026; menus, limits and prices change, so confirm details on the linked pages.
Before you start
- A plan, draft or decision you are unsure about
- Two chats you can open separately
Review evidence
Documentation-based review · 2026-10-24
Primary documentation and official guidance checked on 24 October 2026. Examples, checks and exercises are original illustrations, not recorded tool tests. Cover artwork is a conceptual illustration.
Ask without revealing what you hope to hear
Remove phrases such as “I think this is great, but…” or “my boss loves it”. Present the plan as neutral facts and ask for strengths and weaknesses in equal number. Your opinion is part of the prompt, and the chatbot will notice it.
Check your result
The prompt contains no hint of your preference and asks for equal numbers of strengths and weaknesses.
Ask for the strongest case against it
In a second message, ask for the best argument a smart sceptic would make, the three assumptions most likely to be wrong and the evidence that would change the answer. Insist on concrete reasons rather than polite hedging.
Act as a sceptical reviewer who wants to find the flaw. Here is a plan: [plan]. Give me: 1) the strongest argument against it, 2) the three assumptions most likely to be wrong, 3) what evidence would change your mind, and 4) one cheap test I could run this week. Do not soften the criticism.Example / expected result
Illustrative example; not a recorded model result.
“Assumption 2: customers will pay monthly. Evidence that would change my mind: five paid pre-orders at the proposed price.”
Check your result
You get three named assumptions and a list of evidence that would prove each one wrong.
Get a blind second opinion in a fresh chat
Open a new chat and present the same plan as if a colleague wrote it, with no history and no hint of your view. Compare the two critiques. Points that appear in both are the ones to take seriously.
Check your result
You have two critiques written without your enthusiasm, and a short list of points that appear in both.
When it goes wrong
An assistant forced to criticise may invent weak objections, so judge each one on evidence. A critique is not a decision, and you still need real customer or expert input. Do not share private business details with tools you have not checked.
Make it your own
Take a decision you are leaning toward and run all three steps. Write down the one criticism you did not want to hear and one action you will take because of it.
Useful next resources
Sources and review
Last checked Oct 24, 2026. Features and account limits can change. Verify important results against your original material.
Did this help you complete your task?
Tell us what needs improving. No account or private output required.
Report incorrect information or a broken linkGenius Forges Editorial
Practical AI lessons with checked sources, labelled examples and human verification.
