Nearly half of applicants now ask AI to help choose a university. One in five has crossed a name off because of the answer.

In February, EAB published a survey of more than 5,000 American high-school students. 46% had used AI tools such as ChatGPT in their college search, up from 26% less than a year earlier. Nearly one in five had removed a college from their list because of what an AI-generated answer told them.

That last figure is the one I keep coming back to. It describes a rejection no university saw happen. There was no enquiry, no open day booking and no website visit to show up in the analytics. The student simply never arrived.

Search itself is moving the same way. Pew Research Center tracked almost 69,000 real Google searches last year and found people clicked a result in 8% of visits when an AI summary appeared, against 15% when it didn’t. The answer is increasingly read on the results page, not on your website.

Why the shortlist has become a reputation issue

When I wrote about AI search earlier this month, the question was what these tools say about your university. The harder question is what happens next. A family asks for the best universities for nursing in the north of England, reads a confident paragraph and builds a shortlist from it. If you’re missing from that paragraph, or you appear with an out-of-date fee or a stale statistic, you lose them before your marketing has had a chance to work.

Most of our measurement can’t see this. Website sessions, enquiry volumes and conversion rates all start after the moment that now matters most. A university can report healthy engagement while quietly disappearing from the lists families actually use.

The most important shortlist in recruitment is now written by a machine, and most universities have no idea whether they’re on it.

What I’d ask every director of marketing and recruitment to do

Ask applicants how they built their shortlist. Add a simple question to applicant and enrolment surveys: did you use an AI tool, and what did it tell you? Ask the people who declined your offer as well. It’s the cheapest insight available, and very few universities are collecting it.

Test the questions families really ask. Each term, run the prompts a parent or sixth former would use, by subject, region and price point, across the main assistants. Record whether you appear, how you’re described and which sources are cited.

Retire website sessions as a headline measure. Falling traffic may simply mean the answer was read elsewhere. Report presence and accuracy in AI answers alongside enquiries, applications and conversion, so the board sees the whole funnel.

Publish one clear source of facts. Fees, entry requirements, outcomes, accreditation and campus details, kept current on a single, well-structured page for each course. AI tools reward clarity and consistency, and they punish contradictions between your prospectus, your website and third-party listings.

Earn the citations. Assistants lean heavily on independent sources. Accurate entries in guides, credible coverage and well-maintained listings now shape the shortlist as much as your own pages do.

The opportunity

Few universities are measuring this yet, so the ones that start now will see the invisible shortlist before their competitors do, and recruit from it while others wonder where demand went.

I’ll be asking recruitment leaders on Season 2 of the podcast how they’re tracking it, because the institutions that solve this measurement problem first will set the standard for the rest of the sector.

Jonathan Sabarre FCIM MBA is Principal Consultant at QS Quacquarelli Symonds, a keynote speaker and host of Marketing Greats in Higher Education. Get the weekly email or invite him to speak.