Proofreading Prompt

What it does

Fixes the mechanical errors in your writing — spelling, grammar, punctuation, awkward phrasing — while leaving your actual voice alone. The trick with a proofreading prompt is restraint. Ask an AI to "improve" your text and it'll flatten everything into the same polished-but-forgettable register. This one is scoped tightly: correct what's wrong, change nothing that's merely your style.

Proofread the text below. Fix spelling, grammar, punctuation, and clearly awkward phrasing only.

Rules:
- Do not rewrite sentences that are already correct, even if you'd phrase them differently.
- Keep my voice, word choices, and sentence rhythm. Do not make it more formal or more "professional" unless something is actually wrong.
- Do not add or remove content. No new sentences, no deleted points.
- If a sentence is genuinely unclear, fix it minimally and add a short note in brackets explaining what you changed.

After the corrected version, give me a short list of the kinds of mistakes I made most often (e.g., comma splices, its/it's) so I can watch for them next time.

Text to proofread:
[paste your text here]

How to use it

Paste the full text — an email, a blog post, a cover letter, a chapter. The output comes back in two parts: the cleaned copy, and a short list of your recurring mistakes. That second part is the underrated half. Most people make the same three or four errors over and over, and seeing them named is how you stop making them.

If you write in a specific dialect, say so up front — "use British spelling" or "keep American conventions." Otherwise the model defaults to American English and may "correct" colour to color.

Example output

Corrected text: Your writing, returned with errors fixed and nothing else touched. A comma added here, a "their" swapped for "there," a run-on split into two.

Error summary: A few lines like "You used several comma splices — joining two full sentences with a comma instead of a period or semicolon" and "Watch its (possessive) vs. it's (it is)."

Variations

Stricter mode: Add "Also flag any sentence longer than 30 words as a readability issue, but don't change it unless I ask."

Track changes style: Add "Show your corrections inline using ~~strikethrough~~ for removed text and bold for added text, so I can see exactly what changed."

Formal document: Add "This is a legal/official document. Be conservative — flag anything ambiguous rather than fixing it, so I can decide."

Non-native writer: Add "English is my second language. In the error summary, briefly explain the rule behind each correction so I learn it, not just fix it."

Common pitfalls

Letting it rewrite. If the output reads smoother but no longer sounds like you, the "do not rewrite" instruction got ignored. Re-run and add: "You rewrote several correct sentences. Only fix actual errors."

Skipping the proof of the proof. AI proofreaders occasionally "correct" something that was right — a proper noun, an intentional fragment, a brand name. Always skim the changes; don't paste back blindly.

One giant block. For anything over ~1,500 words, proofread in sections. Accuracy drops on very long inputs, and the model starts summarizing instead of correcting.

Who uses this prompt

Anyone who writes and doesn't have an editor — freelancers sending client work, small-business owners writing their own emails and web copy, students before they hit submit, non-native English writers who want corrections plus the reasoning. It's the prompt for when you trust your ideas but not your typing.