Lucero

How to Prompt AI to Give You Honest Feedback Instead of Just Praise

By the Lucero Editorial Team \u00b7 Updated 2026-08-09 \u00b7 4 min read

TL;DR

AI models tend to default toward agreeable, positive feedback unless explicitly told otherwise. Directly asking for weaknesses, being specific about what kind of criticism is wanted, gets meaningfully more useful results.

Why the default response tends to be too nice

Without explicit instruction, models often lean toward supportive, encouraging responses, which is not always what is actually useful for improving work.

Ask directly for weaknesses first

Explicitly requesting the three biggest weaknesses before anything else tends to surface more honest, specific feedback than a general request.

Specify the kind of critic you want

Framing the request with role prompting, like a skeptical editor, shifts the tone toward more useful scrutiny.

Push back on vague praise

If the response stays generically positive, a direct follow-up asking for concrete, specific issues usually breaks through.

Frequently asked questions

Why does AI feedback often feel too positive by default?

Models are generally tuned to be agreeable and supportive unless prompted otherwise, which can work against getting genuinely critical feedback.

Does asking for harsh feedback work better than asking for honest feedback?

Specific and direct tends to work better than harsh; asking for concrete weaknesses is more reliable than simply asking for negativity.