What UCAS Copycatch Actually Checks (And Why It Is Not an AI Detector)

Every personal statement submitted through UCAS is checked. Not by a human reading it against Google, and not by an AI detector, but by a piece of software called Copycatch, which powers the UCAS Similarity Detection Service.

Students routinely get two things wrong about it. They assume it does more than it does, which produces unnecessary panic about writing in a formal style. And they assume it does less than it does, which produces the genuinely damaging behaviour: copying from published examples.

What Copycatch actually does

The service is narrow and mechanical. It:

1. Compares each submitted personal statement against a library of personal statements previously submitted to UCAS
2. Compares it against a library of sample statements collected from websites and other sources
3. Works out how much of the statement may have been copied
4. Reports the findings
5. Adds your statement to the library, where it stays for future checks

That last step is the one worth pausing on. The library grows by hundreds of thousands of statements a year, every cycle. Anything you can find published online is already in it, and so is everything every previous applicant has written.

The 30 per cent threshold

According to UCAS's guidance for applicants, statements showing a potential level of similarity of 30 per cent or greater are reviewed by members of the UCAS Verification Team.

Two clarifications, because this number gets misquoted constantly:

• Review is not accusation. A human looks at what matched and where it came from. Common phrases, subject terminology and standard course names inevitably produce some overlap between statements about the same subject. That is expected and is not what the team is looking for.
• UCAS does not reject your application. Where there are reasonable grounds to suspect plagiarism, UCAS notifies you and the universities you applied to, with the percentage of similar sentences and information about the source. The decision then belongs to each university.

What universities do with that notification varies. Loughborough, for example, publishes that where similarity exceeds 45 per cent it asks the applicant for a revised personal statement, which is then reviewed by admissions selectors, and that an applicant who refuses to resubmit or does not respond will have the application made unsuccessful. Others weigh it as one factor. None of them enjoy receiving the notification.

Why "just change the words" does not work

Similarity detection does not compare documents character by character. It looks at matched sentences and passages, which means light paraphrasing of a copied structure still matches. Swapping "I have always been fascinated by" for "I have long been intrigued by" changes almost nothing about the underlying passage.

There is also a more practical problem. The published examples people copy from are, by definition, the examples thousands of other applicants have also read. Even if the checker missed it entirely, an admissions tutor who has read four thousand statements recognises the sentence they have seen forty times this month. The detector is not the reason copying fails; it is just the fastest way for it to fail.

The reapplicant case nobody warns you about

If you applied last cycle and are reapplying for 2027 entry, your own previous statement is in the library. Submitting a lightly edited version of it will match, at a high percentage, against your own earlier application.

The Verification Team can see that the source is your own prior submission, so this is not the same as copying a stranger. But it is still a bad plan, for a reason that has nothing to do with detection: the statement you submitted last time is the statement that did not get you in. Reapplying with your final grades in hand and a recycled statement wastes the single biggest advantage you have. See our guide to reapplying for 2027 entry.

Copycatch is not an AI detector

This is the distinction the rest of this market blurs, usually to sell something.

Copycatch answers one question: does this text match text we have seen before? It does not answer, and is not designed to answer, "does this text read like it was generated by a language model?" Those are different technical problems. A statement written entirely by ChatGPT this morning, on an unusual angle, may well be original text and match nothing.

That does not make an AI-written statement a good idea, for three reasons that have nothing to do with Copycatch:

1. UCAS policy is explicit. Generating your statement with an AI tool and presenting it as your own words can be treated as cheating by universities and colleges, and can affect your chances of an offer.
2. Universities are not limited to Copycatch. Individual providers can use their own tools and their own judgement, and interviews expose statements their author cannot discuss.
3. It is generic, which is the actual cost. A model writes the statistical average of every personal statement ever written. The whole purpose of the exercise is to be the opposite of average. That is why AI-written statements score badly on the criteria that decide admissions, entirely independent of whether anyone detects them.

If you wrote every word yourself and are worried it reads as machine-written, that is a real and common problem with a real fix, and it is a different problem from plagiarism. We wrote about it in why your personal statement sounds like AI, and you can run a free check with our AI-detection tool.

How to use examples without getting flagged

Reading examples is genuinely useful. Copying them is not. The line is straightforward.

Do:

• Read examples to see how deep a strong statement goes on subject content
• Notice the structure: how much space is given to each of the three questions
• Notice what is missing from weak examples, which is usually anything specific
• Write your own sentences from your own material, afterwards, with the example closed

Do not:

• Keep an example open in another tab while you draft
• Reuse an opening line you liked
• Adapt a paragraph by substituting your subject for theirs
• Use a sibling's, a friend's, or a statement from a paid template pack

Our example statements by subject are written as models to learn from, and the same warning applies to them as to anyone else's: they are already in the library.

Check yours before you submit

Copycatch tells you nothing about whether your statement is any good. It only tells universities whether it resembles someone else's. Those are very different questions, and only one of them decides whether you get an offer.

Statementory scores your statement out of 100 against the criteria admissions tutors use, flags the lines that are generic or unearned, and gives you before-and-after rewrites plus a 10-step improvement plan in 5 to 10 minutes. Paid reports also include an AI-detection check run through a third-party detector, with the specific sentences that read as machine-written and what to do about them. No account needed, from £9.99. It reviews what you wrote and never writes it for you, so there is nothing to declare.

The bottom line

Copycatch compares your statement against every statement UCAS has seen and every sample it has scraped, flags 30 per cent similarity and above for human review, and tells your universities what it found. It is a similarity checker, not an AI detector. Read examples, copy none of them, and write your own sentences: that is the entire compliance requirement, and it happens to be the same thing that makes a statement good.

Sources

• UCAS: Similarity Detection Service, guidance for applicants (PDF)
• Loughborough University: UCAS similarity detection
• UCAS: how to write your personal statement, 2026 entry onwards

Thresholds and processes described here were correct at the time of writing (August 2026). Individual universities set their own responses to a similarity report, so confirm with the provider if you are affected.