# Feature Request: Prompt Coach / Critical Thinking Mode
I’d like to suggest an optional ChatGPT feature that evaluates the quality of a user’s prompt, not just the quality of the answer.
AI can be influenced by leading questions, confirmation bias and sycophancy. For example:
“Why are climate scientists exaggerating the effect of CO₂?”
already assumes that exaggeration is happening.
A good AI should recognise that assumption rather than simply accepting it. But at the moment, the responsibility often falls on the user to remember to say things like:
“Challenge my assumptions.”
“Don’t reinforce my confirmation bias.”
“Give me evidence for and against my position.”
The users who most need those safeguards may be the least likely to know they should ask for them.
I’d therefore suggest a setting such as **Prompt Coach**, **Prompt Evaluation**, or **Critical Thinking Mode**.
When enabled, ChatGPT could briefly flag issues such as:
* leading or loaded assumptions;
* confirmation-seeking wording;
* false dichotomies;
* claims disguised as questions;
* missing scope or ambiguity;
* weak or anecdotal evidence;
* unfalsifiable framing.
For example:
**Prompt:**
“Why are climate scientists exaggerating the effect of CO₂?”
**Prompt Coach:**
“This question assumes exaggeration has already been established. A more neutral formulation would be: ‘What does the evidence show about the effect of CO₂ on current climate change, and where are the main uncertainties?’”
ChatGPT could then answer the improved underlying question.
I think the real value would be educational. Repeated feedback could teach users to recognise biased framing, unsupported assumptions, poor evidence, missing alternative explanations and questions that cannot be falsified.
Over time, users should become better at asking good questions without needing AI to correct them.
That could actually reduce dependence on AI.
Instead of simply producing polished answers, AI could help users improve their own reasoning.
Different levels might work well:
**Normal** — ChatGPT silently compensates for misleading assumptions.
**Prompt Coach** — briefly explains important weaknesses in the prompt and suggests a better formulation.
**Critical Thinking Mode** — actively tests assumptions, looks for counter-evidence and asks what evidence would falsify the user’s position.
This could also help address sycophancy. AI should help users reach well-founded conclusions, not merely validate beliefs they already hold.
A useful guiding principle might be:
**User preference should influence style, but not reality.**
There is a lot of discussion about making AI more intelligent. An equally important
question is:
**Can using AI make its users better thinkers?**