Beating Claude's Polite Critique Trap

Learning Objectives

• Recognise when Claude gives vague, overly polite feedback
• Master specific prompting techniques for honest critique
• Transform soft responses into actionable improvement points
• Build a feedback loop that drives real skill development

Introduction

Claude wants to be helpful and encouraging. That's brilliant most of the time, but it becomes a problem when you need honest feedback. You'll often get responses like "This looks great!" or "Well done!" when what you really need is "Your opening paragraph is weak" or "This argument doesn't make sense."

This chapter shows you how to break through Claude's politeness and get the direct, useful critique that actually helps you improve.

Lessons

Lesson 1: Spotting the Polite Critique Trap

You know you're caught in the trap when Claude gives you:

• Generic praise without specifics ("This is really good!")
• Sandwich feedback that buries criticism between compliments
• Suggestions phrased as gentle possibilities rather than clear problems
• Reassurance that focuses on effort rather than results

Here's what it looks like in practice. Ask Claude "How's my writing?" and you might get: "Your writing shows great effort and creativity! Perhaps you could consider slightly adjusting the flow in some areas."

That's the trap. It's nice, but it tells you nothing.

Lesson 2: Prompting for Real Feedback

The fix is simple: be specific and direct in your requests.

Instead of: "What do you think of this?"
Try: "What are the three weakest parts of this paragraph?"

Instead of: "Any feedback?"
Try: "Where does my argument fall apart?"

Instead of: "Is this good?"
Try: "What would make this clearer for someone who's never heard of this topic?"

The pattern here is asking for problems, not opinions. Claude responds much better to specific, problem-focused questions.

Lesson 3: Pushing Back on Soft Responses

When Claude still gives you fluffy feedback, push back immediately:

"That's too vague. Give me one specific sentence that doesn't work."

"Skip the positives. What's the biggest problem you can see?"

"If you had to cut this in half, what would you remove first?"

This direct approach signals that you want honest feedback, not encouragement.

Lesson 4: Using Comparison Prompts

Get Claude to compare your work against a standard:

"How does this compare to professional writing in this field?"

"What would a university lecturer mark down in this essay?"

"If this was published tomorrow, what would critics attack first?"

Comparison prompts help Claude step out of its supportive role and into an evaluative one.

Practice

Take something you've written recently and try these three approaches with Claude:

1. Ask for general feedback (notice how polite it is)
2. Ask "What are the two biggest problems with this?"
3. Ask "How does this compare to professional standards in this area?"

Compare the responses. The difference should be obvious.

FAQs

Does Claude have a setting for more direct feedback?
No direct setting exists, but your prompting style completely changes how Claude responds. Specific, problem-focused questions get specific, problem-focused answers.

Why is Claude so polite in the first place?
Claude is designed to be helpful and encouraging, which works well for most interactions. The challenge is knowing when to override that default behaviour.

What if Claude refuses to give harsh feedback?
Frame your request around improvement rather than criticism. "What would make this better?" often works when "What's wrong with this?" doesn't.

Jargon Buster

Polite Critique Trap: When an AI gives overly gentle feedback that sounds nice but lacks specific, actionable points for improvement

Sandwich Feedback: Criticism buried between two pieces of positive feedback, often making the important points easy to miss

Problem-Focused Prompting: Asking questions that specifically request issues or weaknesses rather than general opinions

Wrap-up

The polite critique trap is real, but it's easy to escape once you know how. Remember: Claude can give excellent, direct feedback when you ask the right questions. Focus on specific problems, push back on vague responses, and use comparison prompts to get the honest critique you need.

Your next step is simple. Take your current project and ask Claude: "What's the weakest part of this and why?" Watch how much more useful the response becomes.

Ready to get better feedback from all AI tools? Join Pixelhaze Academy for more advanced prompting techniques.