ChatGPT
Claude
Perplexity
Gemini
GrokThe numbers worth quoting, with publisher, date and sample
Applicant use of AI is well documented in 2026; evidence on how admissions teams use it is thin. This roundup keeps only figures we opened at the primary source. Each line gives the publisher, the sample and the date. None comes from Skolbot.
A method note: no paid keyword-volume tool was available for this piece, so the topic was scoped from public search results and primary sources only.
| Figure | Source | Scope and date |
|---|---|---|
| 95% of students use AI in some form; 94% use generative AI for assessed work | HEPI and Kortext | 1,054 full-time UK undergraduates, December 2025 |
| 43% say a university website is their first port of call; 13% would start with a chatbot | UCAS, reported by Wonkhe | 4,485 potential applicants, applicants and first-years, November 2025 |
| 48% would drop a university after a bad open day; 8% after a negative AI review | Same UCAS research | Same sample |
| 46% of high school students use AI to search for colleges (26% in spring 2025) | EAB | 5,000+ US high school students, October-November 2025 |
| 68% of selective US colleges have no policy on AI in admissions essays | Kaplan (press release, 3 November 2025) | Admissions officers at around 220 US colleges, July-August 2025 |
Two of the five are UK data. Two are American, and they are labelled as such throughout.
Applicants: AI is already part of the journey
AI use among UK students is close to universal. HEPI's 2026 survey, run by Savanta in December 2025, found 95% using AI in some form and 94% using generative AI to help with assessed work. A year earlier, HEPI reported 92% for any AI tool, up from 66% in 2024.
The same 2026 report found that only 36% of students feel encouraged by their institution to use AI, and 38% say their institution provides tools. These are enrolled undergraduates, not applicants. They still indicate the habits sixth-formers will bring with them.
Across Europe, Eurostat found that 33% of people aged 16-74 in the EU used generative AI in the previous three months in 2025, and 64% of those aged 16-24. The UK is not part of that dataset, so treat it as context for competing EU institutions, not as a UK figure.
Research stage: AI matters, but your website still comes first
The UCAS evidence is the most useful for a UK recruitment team. In a Wonkhe article by UCAS's Senior Insight Lead, the November 2025 survey of 4,485 people shows university websites ahead of chatbots as the first stop (43% against 13%).
It also shows where the weight falls. Nearly half (48%) would remove a university from a shortlist after a bad open day, but only 8% would do so after a negative AI review. In other words, the lived experience of the institution still outweighs what an AI says about it.
In the US, EAB's survey shows faster adoption: 46% of high school students use AI in their college search, up from 26% in spring 2025, and 18% have removed a college from consideration because of information surfaced by AI search. Different country, different population, different question: do not add these numbers to the UCAS ones.
Institutions: few data points, many rules still unwritten
The only institution-side figure we could verify concerns the rules set for applicants. Kaplan (press release, 3 November 2025) polled admissions officers at selective US colleges in July and August 2025. Two per cent allow generative AI to write an admissions essay, 30% ban it and 68% have no policy. For brainstorming, 27% allow it, up from 18% in 2024.
This is a US sample, and the UK has its own structure: a personal statement through UCAS, Clearing, and course-level admissions decisions. We found no UK survey of admissions offices that we could open and cite, so we do not quote one. Circulating claims about "the share of universities using AI to screen applications" come from surveys mixing schools and universities, and we could not trace a primary source.
Regulation: what applies to a UK admissions team
For a UK institution, the UK GDPR and the Data Protection Act 2018 are the baseline, supervised by the Information Commissioner's Office, which publishes guidance on AI and data protection. Automated decisions with significant effects on individuals carry specific safeguards, so an admission decision should keep a human in the loop.
The EU AI Act does not apply to the UK as such. Its Annex III lists AI systems that determine access or admission to educational institutions as high-risk, and the European Commission states that Annex III rules now apply from 2 December 2027. It matters if you recruit EU applicants with tools placed on the EU market, and it shows where regulators are heading.
For quality and consumer rules, the Office for Students and the CMA's consumer-law guidance still frame what you can say to prospective students. An AI assistant that gives wrong fee or course information is a consumer-law risk, not only a service issue.
What these figures do not tell you
They do not show that a chatbot raises enrolments. None of the sources measures conversion. They are not directly comparable: students versus applicants, UK versus US, different questions and dates. And they do not measure UK admissions-office adoption.
For published conversion benchmarks, see our article on school website conversion rate benchmarks.
What an admissions team can do with them
Four actions follow without relying on any invented number.
- Make your course pages the best answer to the questions applicants already ask an AI: entry requirements, fees, accommodation, deadlines, Clearing.
- Publish a clear position on AI in personal statements and applications, since most institutions in the Kaplan sample have not.
- Keep decisions with people. Use AI for answering and routing questions.
- Measure your own enquiries, response times and booked open-day places before borrowing benchmarks.
Further reading: our guide to AI agents for student recruitment, the definition of an AI admissions agent and the AI chatbot student recruitment guide.
