ChatGPT Prompts for College Essay Feedback (Not Rewrites)

Start every session with the ground rules

Paste this before your essay, every single time. Chatbots drift back toward rewriting as a conversation gets longer, and restating the rule at the top of a fresh session is easier than catching the drift later.

“You're reviewing a college application essay I wrote. Don't rewrite any sentence, suggest replacement wording, or give example lines, even if I ask later. Every comment must quote the exact words from my draft that it's about. If you don't have a specific comment, say so instead of inventing one.”

That last sentence matters more than it looks. Tell a model to find problems and it'll find problems, whether or not they're there.

Ask for a diagnosis before any line edits

The most useful first pass isn't about sentences at all. It's about whether the essay is doing its job. Paste the prompt you're answering along with the draft, then try these one at a time:

• “In two sentences, tell me what you think this essay is about. Then tell me where, if anywhere, that differs from the prompt.”
• “Which paragraph would a reader remember an hour later, and which one could appear in anyone's essay? Quote the opening words of each.”
• “Where do I state a feeling or a lesson instead of showing the moment that produced it? Quote the line.”
• “What question would a curious reader still have after the last paragraph?”

Turn vague comments into questions you can answer

“Be more specific” is the most common piece of essay feedback, and it's almost never useful on its own. Which sentence? Specific how? Push the model one step further so the comment turns into something you can act on:

“For each place you think is vague, ask me one question about what actually happened that would help me write it more specifically. Don't answer the question for me.” Or, for a single scene: “Ask me three questions about the moment in paragraph two that you'd want answered as a reader. Make them about details only I would know.”

Answer those questions in a separate document, by hand if that helps you think. The answers are raw material. Some of them will end up in the essay, written in your own voice, and plenty won't, which is completely fine.

Test supplements against the question they ask

Supplements fail differently from personal statements. The usual problem isn't a weak scene. It's an answer that could be pasted into another school's application, or one that quietly skips half of a two-part prompt. Both are easy for a model to spot when you ask directly:

• “Here's the prompt and my answer. Which sentences could I have written about almost any college? Quote them.”
• “List each part of this prompt separately. For each one, quote where my answer addresses it, or tell me it doesn't.”
• “Which claim in my answer would a reader at this school want evidence for? Quote it and ask me what the evidence is.”

Have it point at cuts instead of making them

Over the 650-word Common App limit? Asking a model to shorten the essay hands it the pen. Ask it to point instead.

“My draft needs to come in under 650 words. Don't rewrite or shorten anything. List the sentences that repeat an idea made elsewhere, and any places where two sentences say the same thing. Quote each one.” Then you decide what goes, and you do the cutting.

Skip the prompts that quietly become ghostwriting

Some requests sound like feedback and work like ghostwriting. If the reply contains sentences you could paste into your essay, you're close to what Common App's fraud policy describes: submitting the substantive content or output of an AI platform as your own work. These are the usual suspects:

• “Make this paragraph flow better.” The output is a rewrite.
• “Show me what a stronger opening could look like.” An example becomes a template you'll echo.
• “Fix the grammar in my whole essay.” A full pass smooths your voice along with your commas. Ask it to list errors instead, and fix them yourself.
• “Write a conclusion that ties it together.” That's just writing the essay.

Check the policy at every school on your list

Common App's policy allows advice and editing help as long as the final content is yours. Individual colleges add their own rules on top of that, and those rules change between cycles, so read the current guidance for each school you're applying to. Nudge's AI policy reference links the official pages.

None of this is legal advice. It's a way of using a general-purpose tool for the part of the job it's good at.

Decide whether a dedicated tool is easier

Holding those rules in every chat takes discipline, and it's easy to slip when you're tired and the deadline is close. Nudge is built around the same rules with no way to break them: it reads your draft against the prompt, returns five to eight observations tied to exact excerpts, and asks questions while you revise. There's no rewrite mode to slip into.

Use whichever fits how you work. A chatbot with good prompts is a real option, and a free tier often covers it. The thing to protect either way is simple: every sentence you submit came from you!

Primary sources

Rules and prompts can change between application cycles. These are the current first-party sources used to check this guide.

• First-year essay prompts, 2026–2027 — Common App
• Application Fraud Policy — Common App