Common App's fraud policy treats submitting the substantive content or output of an artificial intelligence platform, technology, or algorithm as one's own original work as application fraud. It also allows students to receive advice on content and editing. Those two sentences do most of the work, and they are worth reading before absorbing anyone's summary of them, including this one.
What follows is how that line plays out in practice, what it does not settle, and the test that survives all of it: whether you can account for every sentence in your essay as something you thought and wrote.
Three questions that resolve most cases
- Did a model produce text you submitted?
This is the bright line. Prose generated by a model and then submitted as your own is the thing the policy names, whether you kept it verbatim or edited it afterwards. Editing generated text does not convert it into your writing.
- Could you explain that sentence under questioning?
If a paragraph is in your essay and you could not talk about why it is there, what it refers to, and why you chose those details, it is not functioning as your writing regardless of how it got there.
- What do your specific schools say?
Common App sets a floor, not a ceiling. Individual colleges publish their own guidance, some of it stricter, and they revise it between cycles. The schools on your list are the authority on their own rules.
What tends to be fine, and what tends not to be
These are groupings, not a legal taxonomy, and the schools on your list can draw their lines differently. The pattern underneath them is consistent: help with your thinking and your process sits on one side, and text you did not write sits on the other.
- Understanding what a prompt is actually asking for
- Brainstorming through questions you answer yourself
- Feedback on a draft you wrote, pointing at what is weak
- Proofreading mechanics on finished, student-written text
- Process questions about structure, length, and conventions
- Generating a draft, paragraph, or opening line to submit
- Asking a model to rewrite or improve your prose
- Having a model produce an outline you then fill in
- Letting a model invent details, moments, or reflections
- Submitting anything you could not have written unaided
Why editing generated text does not fix the problem
The common workaround is to generate a draft and then rewrite it heavily, on the theory that enough editing makes it yours. This misreads what the policy is protecting. The application is asking what you thought, and the thinking in that essay was done by the model before you touched it. Editing changes the words, not the authorship of the ideas.
There is a practical version of the same problem. Essays built this way tend to be structurally sound and specifically empty, because a model can only invent the kind of detail that sounds like detail. The result reads as competent and forgettable, which is the outcome the essay exists to avoid.
The risk that is not detection
Most students worry about being caught. That is the smaller risk and the one you have least control over. Detection tools are unreliable in both directions, and a false positive against a student who writes plainly is a real and unfair harm.
The larger risk is that the essay simply does not work. A reader going through hundreds of applications is looking for the one that could only have been written by one person. Generated prose fails that test on its own merits, without anyone needing to prove anything about how it was produced.
How to protect yourself if authorship is ever questioned
The defensible position is a visible process, not a denial. A few habits make that position easy to hold.
- Keep a draft history so the development of the essay is visible over time
- Be able to speak for two minutes about any paragraph without notes
- Write down where each specific detail came from while you still remember
- Check each school's current AI guidance rather than assuming last year's
- Prefer tools that cannot produce submittable text in the first place
Where Nudge sits
Nudge is a question-only coach. It reads your draft, anchors each observation to an exact excerpt from it, and asks you about the gap. It has no mode that drafts, rewrites, outlines, or completes prose, so there is no generated text to submit and nothing to disclose about the words in your essay.
That is a design constraint you can verify rather than a promise you have to take on faith, and it is the specific reason the authorship question does not arise. It is not, however, a substitute for reading the current policies of the schools you are applying to. This page is information, not legal advice.
Common questions
Is using AI for the Common App essay allowed?+
Common App's fraud policy treats submitting AI-generated substantive content as your own original work as application fraud, while advice and editing help are permitted. Individual colleges add their own rules and change them between cycles, so check each school on your list.
Does Common App check essays for AI?+
Common App publishes a fraud policy rather than a description of detection technology, and reliable public detail about what any given admissions office does is limited. Treat the policy as the rule you are held to, and do not treat the uncertainty about detection as permission.
Can I use AI to brainstorm my Common App essay?+
Brainstorming where you answer the questions and write the material yourself is generally treated as ordinary help, much like talking it through with a teacher. It becomes a problem when the model supplies the ideas or the language you then submit.
Is it safe to use AI for grammar and spelling?+
Mechanical proofreading of text you wrote is the least contentious use, and Common App's own language allows help with readability. The caution is that many tools labelled proofreaders will quietly rewrite sentences, which is a different thing.
What should I do if a school's policy is stricter than Common App's?+
Follow the stricter one. Common App's policy is a floor for the shared application, not a ceiling on what individual colleges may require.