Cambridge Has Decided Where AI Has to Leave the Room

For all the noise around AI and university admissions, Cambridge has landed on a surprisingly sensible position.

It has not banned the technology. It has decided where the technology has to stop.

According to Cambridge’s current guidance on AI and undergraduate applications, applicants may use AI tools as prompts for ideas or to assist with research while preparing a personal statement.

The boundaries have become much harder - If AI generates false claims about an applicant’s personal achievements, Cambridge says those claims will be treated like other forms of application fraud. And once the interview begins, the ambiguity disappears entirely.

AI use is prohibited.

Cambridge also says interviewers may use various methods to randomly check that candidates are not using AI and that serious concerns about whether an interview represented the applicant’s own capabilities may put an application at a disadvantage.

That last part is more important than another debate about whether ChatGPT can help edit a personal statement.

Because Cambridge is making a distinction admissions has always cared about but AI has made much harder to observe: the difference between assistance and ownership.

Applications have never been produced in isolation

There is a slightly fictional idea in admissions that the strongest application is created entirely by the student, untouched by anybody else. That has never really been true: Teachers comment > Parents react > Counsellors question > Friends read > Spellcheck corrects > Search engines help with research > Students discuss their interests with people who know more than they do.

The fact that an application has received assistance is not new. What generative AI changes is the distance between assistance and substitution. The progression can happen almost invisibly.

A student asks: Where is my argument unclear?

Then: Can you make this paragraph clearer?

Then: Can you make it more compelling?

Then: Here are my experiences. Which one gives me the strongest story?

Then eventually: What do these experiences say about me?

Every prompt may feel like a small step. But the intellectual transaction changes dramatically. The first asks technology to examine the expression. The last asks technology to help interpret the person.

Cambridge’s rules implicitly recognize that difference - Using AI as a prompt for ideas is not treated in the same way as submitting fabricated achievements. And neither is treated in the same way as using AI during the interview itself.

The interview becomes more valuable

Generative AI has lowered the cost of producing competent prose. That does not mean every AI-assisted essay is dishonest. It does mean polished writing is becoming a weaker proxy for independent thinking.

A student can now improve grammar, structure, clarity, transitions and tone much faster than before. The final document may become significantly better without the student’s underlying understanding becoming significantly deeper. That creates a new admissions problem.

How do you distinguish between a strong application and a strong applicant?

Cambridge already possesses one answer - The Interview.

Cambridge explains that interviews provide further information about an applicant’s suitability to study at the university. That is precisely why it prohibits AI use during them. The interview converts a polished claim into a live intellectual test.

For applicants preparing for Cambridge, the better standard is therefore not: “Do I have the perfect answer?”

It is: “Can I continue thinking after the prepared answer runs out?”

AI may increase the value of live evaluation

This is larger than Cambridge. Generative AI makes many kinds of finished output easier to produce: written statements, research summaries, presentations, project descriptions even technically impressive portfolios.

As the quality of finished output becomes easier to manufacture, institutions may logically place greater value on evidence of process: Live conversation, oral defence, follow-up questioning, problem-solving, draft history, discussion of sources and the ability to explain why a decision was made.

This does not mean every university will suddenly introduce Cambridge-style interviews.

Interviews are expensive. They can introduce their own biases. They are difficult to scale. But the underlying principle is likely to become more important: when output becomes easier to generate, evidence of independent thinking becomes more valuable.

Cambridge’s policy is one early, visible version of that shift.

The strongest application should survive closing of the laptop

For families, counsellors and students, this gives us a more useful standard than endlessly debating whether a particular AI tool is “allowed.”

Cambridge itself already distinguishes among uses. Limited research and idea prompting may be acceptable. Invented achievements aren't. AI assistance during interviews is explicitly prohibited.

So the important question becomes less technical: If every AI tool disappeared tomorrow, would the applicant still understand, believe and be able to defend what the application says?

If yes - technology probably supported the process.

If no - somewhere along the way the application may have become more capable than the applicant.

Cambridge has not banned AI. It has simply identified the moment when the machine has to leave and the student has to remain. That may prove to be one of the more important admissions distinctions of the AI era.