You wrote every word of it and it still reads like a machine produced it. That is extremely common, it is not a rule breach, and the reason it happens is fixable in an afternoon.
The problem is almost never that you cheated. It is that you wrote towards what you thought a personal statement was meant to sound like.
Check my statement, £9.99 →Includes an independent AI-detection check · Ready in 5–10 minutes · No account needed
A language model learned to write personal statements by reading an enormous number of them, along with everything else in the same register: application advice, prospectus copy, motivational writing. When it produces a statement, it produces the average of that material.
When you sat down to write yours, you did something remarkably similar. You read a few examples, worked out what the genre sounded like, and wrote towards it. Careful, formal, balanced, ambitious about impact. That is a sensible strategy and it is why your statement now collides with the machine average.
So the sentences that read as generated are usually the ones where you were trying hardest to sound like a proper applicant. That is worth knowing, because it tells you exactly where to look: not at your typos, at your best behaviour.
None of these are about vocabulary. They are all about a sentence making a claim it never pays for.
“I have always been passionate about justice and fairness.”
Passion is asserted, never evidenced. Any applicant to any law course could have written it, which is precisely what makes it read as generated: a model producing the average sentence for this genre produces this sentence.
“This experience gave me valuable transferable skills.”
Experience, valuable, transferable, skills. Four abstractions and not one concrete thing. Strip the nouns out and there is no information left in the sentence at all.
“Furthermore, this deepened my understanding of the subject.”
Real thinking is uneven. It doubles back, it hesitates, it gets stuck on one detail. Prose that glides from point to point with Furthermore and Moreover reads as assembled rather than thought.
“Reading this book taught me a lot about the field.”
The book is named but never argued with. “This taught me” is a report. “This changed how I think about X, and here is what I still cannot resolve” is a position, and only you can hold one.
Reaching for a thesaurus keeps the hollow structure and changes the paint. Replace the claim with something only you could have written instead.
Interchangeable
“I have always been fascinated by the human body.”
Only you could write this
“I did not expect the heart to be a pump that adjusts its own output. Reading about the Frank-Starling mechanism, I got stuck on the fact that a stretched ventricle contracts harder without any signal from the brain.”
One named mechanism, one thing the writer did not expect. Nobody else has that sentence.
Interchangeable
“Volunteering taught me the importance of communication.”
Only you could write this
“The tenant in that possession hearing did not know she could ask for an adjournment. Everyone in the room except her knew what was going to happen in the next eleven minutes.”
The lesson is not stated. It is shown, and it is far harder to forget.
Interchangeable
“I enjoy the problem-solving aspect of mathematics.”
Only you could write this
“I spent most of a half-term convinced I had found a counterexample to the pigeonhole principle. I had not. Finding my own error taught me more than the proof did.”
A specific failure, a specific length of time. Generic prose never admits to being wrong.
Notice what changed: not the vocabulary, the amount of the world in the sentence. A named mechanism, a length of time, a mistake, a number.
UCAS does not run an AI detector on your personal statement. It runs Copycatch, a similarity check that compares your statement against previously submitted statements and published sources. It is looking for copied text, not for machine-generated text.
So if your worry is that a detector is going to flag you and sink your application: it will not. A statement that reads as generic is not a procedural risk, and anyone telling you otherwise is selling something.
The cost is real but it is a different one. An admissions tutor works through hundreds of statements for the same course. Yours does not get rejected for sounding generic. It gets forgotten, which in a competitive course is the same outcome by a slower route.
More on this: do universities check for AI in personal statements and what happens if UCAS detects AI.
Every paid review includes an AI-detection check. Here is exactly what that means, and what it deliberately does not do.
Your statement is run through an independent third-party AI-detection service. Not our opinion of your writing style: a tool built to separate human from machine text, scoring every sentence separately.
Detectors have real false-positive rates, especially with students who write carefully or in a second language. Any flagged line that contains something only you could have written is dropped from the report, and we say which ones and why.
For every line that survives, the report shows what makes it interchangeable and asks you one question only you can answer, pointing at material already elsewhere in your own statement. It never writes the new sentence for you.
Why we will not rewrite the line for you. Handing you a machine-written sentence to replace a sentence that reads as machine-written solves nothing, and submitting words you did not write is the one thing UCAS actually prohibits. Statementory checks what you wrote. It has never written a statement for anybody.
Want to test the water first? The free AI-style checker gives you our own read on the writing in seconds.
Because you almost certainly wrote it towards what you thought a personal statement was supposed to sound like. Formal register, balanced transitions, big abstract claims about passion and impact: those are the patterns language models learned from the same source material, so writing that imitates the genre reads like writing that was generated. It is a sign you were being careful, not a sign you cheated.
No. UCAS runs Copycatch, a similarity check that compares your statement against previously submitted statements and published sources. It looks for copied text, not for machine-generated text. So a statement that reads as generic is not a procedural risk to your application. The real cost is different: an admissions tutor who reads hundreds of statements for the same course cannot tell your generic sentences apart from anyone else's.
No, and it usually makes it worse. Swapping words keeps the underlying structure, which is what actually reads as generated: a claim with nothing underneath it. The fix is not different words for the same empty sentence, it is replacing the sentence with something only you could have written. A named text, a specific moment, a number, a question you could not answer.
That is the trap. If you ask a model to rewrite a machine-sounding line, you get another machine-written line, usually with the same hollow structure in fresher clothes. It also puts you on the wrong side of the UCAS rules, which prohibit submitting work you did not write. The sentence has to come from you. What a tool can usefully do is show you which lines are the problem and ask you the right question about each one.
Every paid review is run through an independent third-party AI-detection service that scores your statement line by line. The report shows you the exact lines that read as machine-written, explains what makes each one interchangeable, and asks you a question that only you can answer so you can rewrite it in your own words. It never writes the replacement sentence for you, which would defeat the point.
If your application has gone, you cannot edit the statement. It is worth remembering that a generic statement is not a rule breach and will not get your application rejected on its own. If you are applying through Clearing, reapplying next cycle, or still drafting, the fix is entirely available to you and it is mostly a matter of trading claims for specifics.
A full review scores your statement out of 100, annotates it line by line, and runs the independent AI-detection check that tells you exactly which sentences could have been written by anyone.
Get my review, £9.99 →What is in the report · Free AI-style checker · Free preview