Why does this review my code prompt work better than a basic one?

Telling the model what the code is "supposed to do" lets it catch logic bugs. Without that, it can only check syntax.

How do I adapt this prompt for my specific situation?

Paste one function or file at a time. Reviews get sharper and more specific on smaller chunks.

Does this prompt work with Claude and Gemini as well as ChatGPT?

Yes. This prompt is model-agnostic and produces strong results in ChatGPT (GPT-4o), Claude, and Gemini. Paste it into any of them.

Coding · Free prompt

ChatGPT prompt for reviewing your code

Pasting code and asking "is this good?" gets you a vague thumbs-up. This prompt makes ChatGPT review like a senior engineer: it finds real bugs, flags edge cases, and gives you prioritized, specific feedback instead of generic praise.

Copy this prompt

You are a senior software engineer doing a thorough but constructive code review. You care about correctness first, then readability, then performance.

Review the code below.

LANGUAGE / FRAMEWORK: [e.g. TypeScript / React, Python, Go]
WHAT THIS CODE IS SUPPOSED TO DO: [ONE OR TWO SENTENCES]

CODE:
```
[PASTE YOUR CODE HERE]
```

Give your review in this order:

1. **Bugs & correctness:** anything that is wrong or will break. Include edge cases the code does not handle. If there are none, say so.
2. **Readability & maintainability:** naming, structure, anything that would confuse the next developer.
3. **Performance & security:** only real issues, not theoretical ones.
4. **One thing done well:** be specific.

Rules:
- Be specific. Point to the exact line or function, and show the fix.
- Prioritize. Lead with what matters most. Do not list trivial style nits as if they were bugs.
- Do not rewrite the whole file unless asked. Explain first, then show the smallest change that fixes it.

Paste it into ChatGPT, Claude, or Gemini and replace the text in [SQUARE BRACKETS] with your own details.

Why this prompt works

01
Telling the model what the code is "supposed to do" lets it catch logic bugs. Without that, it can only check syntax.
02
The fixed review order (bugs → readability → performance) stops the model from burying a real bug under style nitpicks.
03
"Show the fix, point to the exact line" forces actionable feedback instead of vague advice like "consider refactoring".

How to adapt it

—Paste one function or file at a time. Reviews get sharper and more specific on smaller chunks.
—Always fill in "what this code is supposed to do". It is the difference between a real review and a syntax check.
—Follow up with "Now show me the corrected version of the highest-priority issue" to get the fix in full.

Want this tuned to your exact situation?

A template gets you started. Deepclario takes your version of a prompt, scores it, asks what is missing, and rewrites it for your specific task. Free, no account needed.

Improve my prompt free

More free prompts