“Do I Have to Make My College Essay Worse to Prove I Am Human?” Here’s the Answer.
A counselor’s answer to one of the strangest—and most damaging—new fears in college admissions
“Can we make it a little worse?”
The student said it so sincerely that, for a moment, I did not know what she meant.
We had been working on her personal statement for several weeks. The story was hers. The first draft had been unmistakably hers, too: perceptive in some places, confusing in others, funny without trying to be funny, and far too long. We had watched it develop sentence by sentence. We had heard her explain the memory aloud, change her interpretation of it, discard an opening she once loved, and finally recognize what the experience actually meant to her.
Then she uploaded the finished draft to an AI detector.
The result told her that much of the essay was probably generated or polished by artificial intelligence.
So she came into our next meeting with a plan.
She wanted to remove the em dashes. She wanted to replace a few sophisticated words with simpler ones. She wanted to break up a sentence that flowed too smoothly. She wondered whether she should add a fragment, a contraction, or even a small grammatical error.
Not because any of those changes improved the essay.
Because she wanted to prove that a human had written it.
The scene above is a composite of conversations we are now having with students with startling regularity. The identifying details are intentionally omitted, but the question is real:
Do I now have to make my writing worse—or intentionally less polished—to prove that I am human?
The answer is no.
You should not make your writing worse to satisfy an AI detector. You should make your writing more recognizably yours.
Those are very different goals.
No, polished writing is not evidence of cheating
Let us answer the question as directly as possible.
A grammatically clean sentence is not proof of AI use. An em dash is not proof of AI use. A strong metaphor, an elegant transition, a sophisticated word, or a carefully constructed conclusion is not proof that ChatGPT wrote your college essay.
Students were capable of beautiful writing before generative AI existed. They remain capable of it now.
Neither the Common Application’s fraud policy nor the published policies of selective colleges say that an applicant’s writing must be awkward, simplistic, or imperfect. They focus on authorship and ownership. Common App identifies misrepresenting the substantive output of an AI platform as one’s own work as a possible form of fraud. Yale, meanwhile, currently distinguishes between limited uses such as grammar review or early topic suggestions and submitting substantive AI-generated writing as one’s own.
Harvard similarly says that an applicant’s essay should reflect the student’s own content, writing style, and English proficiency. Princeton’s admissions guidance emphasizes writing in a style that reflects the applicant’s own voice and certifying that the application is the student’s own work.
Notice what these institutions are not saying.
They are not saying:
Do not revise.
They are not saying:
Do not write well.
They are not saying:
Make sure your sentences look sufficiently teenage.
They are saying that the thinking, experiences, language, interpretations, and substantive work must honestly belong to the applicant.
The problem is not polish.
The problem is borrowed polish that replaces the student.
Why are strong writers suddenly afraid of sounding strong?
AI has created a strange psychological trap for students.
For years, applicants were told to proofread carefully, vary their sentence structure, eliminate repetition, choose precise words, and revise until the essay flowed naturally. Now many students fear that doing those things too successfully will make their writing appear artificial.
We understand why.
Generative AI can produce clean, organized, superficially reflective prose in seconds. As we explain in What College Admissions Offices Are Looking for in 2026, AI has made generic polish remarkably inexpensive. What remains scarce is not grammatical competence. It is evidence of an individual mind noticing, questioning, connecting, reconsidering, and creating meaning from an experience.
That shift matters.
A sentence can be polished and deeply human.
A sentence can also be polished and completely anonymous.
Consider the difference:
Every Sunday, my grandfather calls at exactly 7:03—not 7:00—because he says punctual people are predictable, but three minutes late means you have a life.
Now compare it with:
Through my relationship with my grandfather, I learned that meaningful connections often emerge through the small rituals that shape our lives.
The second sentence is smoother in a conventional “college essay” way. It also tells us almost nothing.
The first sentence contains a person, a habit, a worldview, a relationship, and a tiny contradiction. It gives the reader something to notice. It makes us wonder about the student’s grandfather and why the student remembers those three minutes.
That is the kind of specificity we discuss in our guide to writing a standout college application essay. The difference is not that one sentence contains errors and the other does not. The difference is that one sentence feels observed while the other feels manufactured.
Human writing is not defined by imperfection. It is defined by ownership.
What happens when a student starts writing for the detector?
This is the part that concerns us most.
A student runs a genuinely human-written essay through one detector. It returns a high AI probability. The student tries another program, which produces a completely different result. Then a third platform offers another percentage.
The student begins revising.
An em dash disappears.
Then a transition.
Then a carefully chosen word.
Then the student adds a random sensory detail—not because it matters to the story, but because someone online claimed that AI does not use enough sensory language.
A natural sentence is interrupted so that its rhythm looks less predictable. A clear reflection is made intentionally clumsier. A phrase the student genuinely uses is deleted because it appeared on a list of “words ChatGPT loves.”
Eventually, the student is no longer asking:
“Does this reveal me?”
The student is asking:
“Will a machine believe me?”
That is exactly the wrong audience.
We examined this problem in detail in Is Your College Admissions Essay Being Falsely Flagged for AI? Here’s What You Should Do. AI detectors do not possess a hidden record of who wrote each sentence. They analyze characteristics of the finished text and generate an estimate. Different systems use different models, thresholds, datasets, and definitions, which is why the same essay may receive dramatically conflicting results.
Even Turnitin’s own guidance acknowledges the possibility of false positives and treats lower-range results as less reliable. Research from Stanford scholars has also found that several detectors disproportionately misclassified writing by non-native English speakers as AI-generated.
A percentage may look scientific.
That does not make it a verdict.
The em dash did not suddenly become dishonest
We should talk about the em dash because students certainly are.
Over the last year, we have heard students worry about punctuation in ways we never encountered before. They ask whether they should remove every em dash because ChatGPT uses them. They worry that parallel structures sound machine-generated. They avoid semicolons. They replace words they naturally know because those words appear on viral lists of supposed “AI vocabulary.”
This is not a sensible way to evaluate authorship.
AI uses commas.
Humans also use commas.
AI writes sentences that begin with “But.”
Humans have been doing that for a very long time.
AI uses em dashes partly because human writers use em dashes—and because large language models were trained on enormous quantities of human language.
No individual punctuation mark proves anything.
The meaningful question is not whether a particular feature sometimes appears in AI-generated writing. The meaningful question is whether the voice, thought, and language are consistent with this student.
A student who has used em dashes throughout years of school writing should not suddenly delete them from a personal statement out of fear. A student who has never used one and cannot explain why six appeared after an AI editing session should look more carefully at how the essay changed.
The punctuation itself is not the issue.
The writing history is.
Polished and over-edited are not the same thing
We are strongly in favor of revision.
A college essay should not necessarily be submitted in the form it took when the student first emptied every thought onto the page at 1:14 in the morning. Revision can clarify the story, strengthen pacing, remove repetition, deepen reflection, and help the reader understand what the student has been trying to say.
But revision should reveal the student more clearly.
It should not replace the student with an idealized applicant.
We explored this distinction in The Over-Edited College Essay: When Adult Help and AI Polish Erase the Student. An essay can become technically stronger while becoming personally weaker. It can contain better vocabulary, cleaner transitions, and a more cinematic conclusion while gradually losing the particular mind that made the story worth reading.
Here is the distinction we use with students:
Healthy polishing makes the student easier to hear.
Unhealthy polishing makes the helper easier to hear.
Healthy feedback may identify that a paragraph is confusing.
Unhealthy editing writes the student a new paragraph.
Healthy feedback may ask why a remembered detail matters.
Unhealthy editing supplies the meaning of that detail.
Healthy feedback may point out that the conclusion sounds generic.
Unhealthy editing writes a profound-sounding lesson the student never actually reached.
A responsible editor, teacher, parent, counselor, or AI tool should help the student communicate—not manufacture a more marketable personality.
That principle also guides our advice in How to Choose a College Essay Writing Consultant Without Losing Your Student’s Voice. The best guidance asks the question that helps a student uncover an answer. It does not hand the student a more impressive answer to submit.
The best test is not “Does this sound like AI?”
During essay meetings, we often ask students to read a difficult sentence aloud.
Not perform it.
Not read it in a formal presentation voice.
Just say it.
Sometimes the student reaches the middle of the sentence, stops, and laughs.
“I would never say that.”
That reaction tells us far more than an AI-detection score.
We then ask the student to look away from the document and explain what the sentence is supposed to mean.
Almost every time, the spoken explanation is better.
It may not be as formally elegant. It may contain a restart or an unexpected phrase. But it is clearer, more specific, and more connected to what the student actually believes.
We write down what the student said.
Then we work from there.
This is why the better question is not:
“Does this sound like AI?”
It is:
“Does this sound like me at my clearest?”
Not you at your most casual.
Not you texting a friend at midnight.
Not an unrevised first draft of you.
You at your clearest, most thoughtful, and most intentional.
A personal statement should sound like a student who had time to think and revise. It just should not sound like a student was replaced by a committee, a parent, a professional copywriter, or a language model.
Five questions that reveal more than any detector
Before changing a strong essay because of an AI score, sit with the document and ask five human questions.
Could I explain every sentence without looking at it?
A student who owns the essay can usually explain why a scene was included, what a metaphor means, why a paragraph moved, and how the conclusion developed.
The explanation does not have to be sophisticated.
It has to be true.
Did I actually reach these conclusions?
AI is very good at attaching familiar lessons to experiences:
I learned the importance of resilience.
Leadership is about empowering others.
I discovered the transformative power of community.
Those ideas are not automatically false. But they are so broad that students may accept them without asking whether they describe what truly changed.
A more revealing conclusion may be smaller, stranger, or less complete.
Perhaps the student did not learn to conquer uncertainty. Perhaps the student learned to stop treating uncertainty as a failure.
That difference belongs to a person.
Are the details present because they matter?
Specificity is not a decoration students sprinkle across an essay to make it “human.”
A detail should reveal something.
The color of a chair matters only when the student remembers the chair for a reason. The smell of a kitchen matters only when it places us inside a meaningful relationship or moment. Random details added to defeat an algorithm make an essay feel more artificial, not less.
Would someone who knows me recognize me?
Give the essay to a trusted teacher, counselor, sibling, or friend without mentioning the detector result.
Do not ask, “Does this sound AI-generated?”
Ask:
“Do you recognize me in this?”
That invites the reader to think about the student rather than speculate about a machine.
Does this voice match the rest of my application?
The personal statement does not exist alone.
Admissions officers also encounter activity descriptions, supplemental essays, short answers, recommendations, academic interests, interviews, and other written materials. Those pieces should not all sound identical, but they should feel compatible.
This is one reason we build applications through App Identity™. A student’s essays, activities, recommendations, intellectual direction, and contribution should offer different chapters of the same person—not introduce a new applicant on every page.
What should you do after a human-written essay is flagged?
First, stop scanning it.
Running the same essay through six detectors will rarely create clarity. It often produces six conflicting results and significantly more anxiety.
Return to the last version you know you substantively wrote. Save your outlines, notes, early drafts, comments, and document version history. Those materials show a developing writing process far more meaningfully than a screenshot displaying an AI percentage.
Then read the essay aloud.
Mark every sentence that feels unlike you. Do not automatically delete it. Ask why it feels wrong. Perhaps someone else rewrote it. Perhaps a grammar tool transformed it more substantially than you realized. Perhaps it is yours, but you are trying too hard to sound profound.
Explain the sentence in your own words and revise from that explanation.
Next, compare the essay with your other materials. Does the humor sound like yours? Do the interests align? Does the level of English proficiency feel consistent? Can the reflections be supported by experiences elsewhere in the application?
Finally, review the published AI policies of the colleges on your list. Do not rely solely on a viral social-media post claiming that “all colleges allow” or “all colleges prohibit” a particular use. Policies are not entirely uniform.
What you should not do is insert mistakes, run the essay through an “AI humanizer,” or distort your natural voice to manipulate an uncertain score.
Revise for truth.
Revise for clarity.
Revise for ownership.
Do not revise for theatrical imperfection.
A special word to multilingual students
Students who learned English as an additional language sometimes feel the greatest pressure to alter their writing.
They may have spent years learning to write carefully, use orderly transitions, avoid idioms they do not fully understand, and construct grammatically consistent sentences. Some detection systems may interpret that linguistic predictability as evidence of generated text.
The Stanford-led research on this issue should make educators and admissions professionals cautious about treating detection scores as proof. The researchers found that several publicly available detectors frequently misclassified writing by non-native English speakers.
The answer is not for multilingual students to make deliberate errors.
Your English proficiency is part of your real academic history. Your writing should reflect your ability—not an exaggerated version of it in either direction.
Do not make yourself sound more advanced than you are.
But do not make yourself sound less capable merely because a machine has made you frightened of competence.
Keep your drafts. Work with people who understand your natural voice. Make sure you can explain every idea and every phrase. Then allow your writing to be as strong as it honestly is.
Parents: please do not help your student “sound human”
Parents are often trying to protect their children.
A detector produces a frightening result, and the family begins editing together. One parent suggests adding contractions. Another recommends removing advanced vocabulary. Someone proposes inserting a joke. Everyone becomes preoccupied with making the essay sound like a believable teenager.
That process can unintentionally create the very problem the family is trying to solve.
The more people manipulate the student’s language, the less naturally the essay belongs to the student.
A parent’s most useful role is not to rewrite. It is to remember.
You can ask:
“Is that how you described it when it happened?”
“What did you actually say to me afterward?”
“Why do you still remember that?”
“You seem more emotional when you talk about this than when you write about it. What is missing?”
“I understand the sentence, but I do not hear you in it. Can you explain it another way?”
Those questions return authorship to the student.
The same principle should govern professional support. Our college admissions services are designed around helping students become more visible throughout the application—not making a counselor visible through the student’s writing.
Why this matters even more now
Some colleges are reducing or removing supplemental essay requirements. As we explain in Colleges Are Dropping Application Essays, but It Is Not Good News for Students, fewer essays mean fewer opportunities for an applicant to reveal different dimensions of identity, intellectual direction, character, humor, and contribution.
That places more responsibility on the writing that remains.
The answer is not to make that writing safer, flatter, or intentionally worse.
It is to make every remaining word more revealing.
Students should also remember that authenticity is larger than AI detection. Colleges are evaluating whether the entire application is credible. Research, nonprofits, businesses, leadership claims, awards, activities, and personal narratives all need to withstand closer examination. Our College Application Authenticity Audit explores how students can evaluate the integrity and consistency of the complete file.
An admissions officer is not merely asking whether a sentence sounds human.
The officer is trying to understand whether the person presented throughout the application is real.
The sentence we want every worried student to hear
You do not need to make your essay worse to prove that you are human.
You do not need to remove every em dash.
You do not need to add a typo.
You do not need to interrupt a beautiful sentence simply because it is beautiful.
You do not need to apologize for being a strong writer.
But you do need to be able to claim the sentence.
You should know where it came from.
You should understand what it means.
You should recognize the thought inside it.
You should be able to explain why it belongs in your story.
And when someone who knows you reads it, that person should not merely say, “This is well written.”
They should say:
“Yes. That is you.”
That is the standard.
Not artificial imperfection.
Not algorithmic approval.
Not a score from a website that has never met you.
Authorship. Specificity. Consistency. Ownership.
Your college essay is not supposed to prove to a machine that you are human.
It is supposed to help another human being understand who you are.
Frequently Asked Questions
Should I add mistakes so my college essay does not sound AI-generated?
No. Intentional errors weaken the essay and do not reliably demonstrate authorship. Preserve your drafts, review the essay for your natural voice, and revise only where the writing is inaccurate, generic, borrowed, or disconnected from how you actually think.
Is a highly polished college essay suspicious?
Not automatically. Thoughtful students revise. The concern arises when the vocabulary, insights, emotional conclusions, or writing level do not appear to belong to the applicant or do not match the rest of the application.
Should I remove em dashes from my Common App essay?
Not merely because AI systems also use them. Punctuation should serve the sentence. Keep an em dash when it reflects your normal writing and improves the passage. Remove it when it is unnecessary—not when social media has made you afraid of it.
What should I do if GPTZero or another detector flags my essay?
Do not panic or immediately rewrite the essay. Save your version history, notes, outlines, and earlier drafts. Read the essay aloud, make sure every substantive idea and phrase is yours, and review each college’s AI policy. Our complete guide to false AI flags in college essays provides a step-by-step response.
Is using AI for grammar correction allowed?
Policies vary. Yale currently states that AI-supported grammar or spelling review does not itself constitute application fraud, while substantive AI-generated content presented as the student’s own writing crosses the line. The difference between correcting punctuation and rewriting an entire paragraph is significant. Always check the rules of each institution.
How can I tell whether an essay still sounds like me?
Read it aloud. Explain every important sentence without looking at the page. Ask someone who knows you whether they recognize your personality and thinking. Compare the voice with your short answers, activity descriptions, school writing, and other application materials.
Can colleges detect ChatGPT in admissions essays?
College policies and review practices vary, and an automated score should not be mistaken for direct proof of authorship. Admissions readers may also notice inconsistencies in voice, specificity, writing ability, or the larger application. Read our broader analysis of whether colleges can detect ChatGPT in admissions essays.
Students and families who are uncertain whether an essay has been over-edited, falsely flagged, or separated from the student’s authentic voice may schedule a private conversation with The Ivy Institute. Our goal is not to make an essay sound more like us. It is to make certain the student has not disappeared from it.
