How to Choose an AI College Essay Tool

Ask whether the AI college essay tool produces prose the student could submit

This is the sharpest dividing line in the category, sharper than price, design, or marketing language. A tool that outputs draft sentences, a rewritten paragraph, a suggested opening line, a polished version of a rough one, puts the authorship question directly on the student, because now something exists that didn't come from them and can be pasted straight in. A tool that only asks questions can't do that, because there's nothing to paste. The two categories carry genuinely different risk, no matter how similar the interfaces look.

So the question worth asking before anything else: if you typed a rough paragraph into this tool right now, would it hand back something you could paste into the application? Common App's fraud policy treats intentionally presenting substantive AI output as one's own original work as application fraud. It doesn't create a universal permission for advice or readability editing, so hold the tool against that boundary and each school's more specific current policy.

Nudge is one data point you can check yourself against this test: it only asks questions about a draft and doesn't generate replacement sentences, a claim you can verify directly by trying it on a real paragraph and watching whether anything resembling finished prose shows up on the screen.

Check whether the feedback is anchored to your child's actual words

Generic feedback is the most common thing sold as personalization in this category. "Consider adding more sensory detail" or "this paragraph could be more specific" is true of nearly any draft on any topic, which means it isn't actually about your child's essay. It's a template response with the student's word count dropped in.

Anchored feedback points to an exact sentence or phrase from the draft and asks something that could only be asked about that sentence. Here's a quick way to test a tool: run the same rough paragraph through it twice, once as written and once with one detail changed, and see whether the feedback changes. Does it? If not, the tool isn't actually reading closely.

Ask what happens to the essay data

A college essay draft often contains a student's name, address, family history, health information, or other personal material well beyond what a typical app collects. Before your student pastes a draft into any tool, ask three plain questions: is this content used to train models? Is it shared with third parties? Can it be deleted on request? A tool's privacy page should answer all three without requiring a support ticket.

Look for a record of how the draft developed

A dated history of a draft, from rough notes through revisions, is what protects a student if authorship is ever questioned, whether by a teacher, a school, or their own later doubt about whether a sentence is really theirs. A tool that only shows the current version and overwrites everything before it gives you no way to reconstruct that story.

This matters independent of whether the tool is well behaved. Even a student working entirely on their own benefits from being able to point to a September draft and a November draft and see the thinking in between.

Decide whether it makes the student think or lets them skip it

This is the criterion that decides whether the essay actually improves or just gets smoother. A tool that surfaces a gap, a missing detail, an unclear motivation, a claim without evidence, and leaves the student to close it is asking for thinking. A tool that closes the gap itself, even under the label of a suggestion the student can accept or reject, is doing the thinking and inviting the student to rubber-stamp it.

Watch for how a tool handles a stuck student. Does it ask a narrower question when the first one doesn't land, or does it just hand over a suggestion? A tool built around thinking tends to do the former. A tool built around output tends to default to the latter, because that's the faster way to keep the session moving. Faster isn't the same as better here.

Understand the price and what it rewards

Pricing structures push behavior even when nobody intends it. A model that charges per generation or per rewrite has a built-in incentive to produce more finished text, because that's the unit being sold. A model priced around feedback sessions or draft reviews has less reason to hand the student a paragraph, because the product is the conversation, not the output.

Ask directly what you're paying for: a number of AI-generated drafts, a number of feedback rounds, unlimited access to one mode of help, or something else. The answer tells you what the tool is optimized to deliver more of.

Ask the student two questions instead of the tool

The most reliable check doesn't involve the tool at all. Ask your student to explain why they changed a specific line, in their own words, without looking anything up. Then pick any paragraph in the essay at random and ask them to talk about it for two minutes without notes: why is it there, what does it mean, why that sentence and not another one?

That two-minute test checks something a privacy policy or feature list can't: whether the writer can explain the essay's choices and reconstruct its central scene. If they can't, that's a problem worth addressing before submission, regardless of which software touched the draft.