A 17-year-old in Berlin or Seoul deciding where to study abroad no longer starts with a brochure. She opens ChatGPT, Claude or Gemini, types “best universities in Europe for business psychology with good job outcomes and a stable visa route”, and gets a shortlist in eight seconds, with no admissions office built and no marketing budget touched.
That shortlist is the real front door of international education now, and most institutions haven’t noticed.
The funnel has a new gatekeeper
For 20 years, international recruitment has been a game of visibility: rank well, show up on page one, get the click. That game assumed a human was clicking. Increasingly, they’re not.
An EAB survey of over 5,000 US high schoolers found 46% now use tools like ChatGPT during their college search, nearly double the 26% recorded in spring 2025, and one in five had already dropped an institution based on what AI told them. INTO University Partnerships’ own survey of 1,600 newly enrolled international students found something more uncomfortable: 96% of enrolled students who used AI said it matched or exceeded official sources.
For 20 years, international recruitment has been a game of visibility: rank well, show up on page one, get the click. That game assumed a human was clicking. Increasingly, they’re not
Marketing teams have spent a decade optimising for Google’s algorithm. Now they need to be legible to a reader that doesn’t click through and forms its answer by synthesising whatever it finds, accurately represented or not. Higher ed marketers are already scrambling to adapt, chasing what’s now called generative or answer engine optimisation. But GEO isn’t a new marketing channel. It’s a new buyer sitting between the institution and the student, and most universities have no idea what it’s been told to say about them.
This isn’t abstract to me. Our own enrollment data this year showed 80% of the EU and US students who enrolled with us first found out we existed through Claude or ChatGPT, not our website or paid campaigns. Credibility has relocated into the AI-mediated layer now sitting between us and the student.
The questions we can’t yet answer
Here’s what unsettles me most: we don’t know why the algorithm is choosing us, and we’re building strategy around it anyway. Is what makes a model treat information as credible thesame discipline SEO demanded, or has the ground moved toward the forums these models lean on, Reddit included?
A second question troubles me too: our AI-driven enquiries skew heavily toward wealthier, English-speaking markets. Are the models steering higher-income users toward private institutions like ours, by design or their training data, or is this an adoption gap that will close over time?
I don’t have a clean answer to either question, and I’d be suspicious of anyone who claims they do. What we do have is a growing pile of decisions, where we invest, which markets we chase, shaped by systems whose logic none of us can fully see. That should trouble this sector more than it does.
The weight of choosing
The same students turning to AI are actively hostile to institutional marketing. EAB found more than half reacted negatively the moment they suspected a college’s message had been written, or polished, by AI. They want machine-verified facts and human-sounding trust, and are getting good at telling the two apart.
That distrust makes sense given what this generation carries: told that platforms and data would make the right choice obvious, and finding neither delivers. Faced with too many options and too little certainty, students either get overwhelmed or hand their judgement to the algorithm, not for convenience but to avoid the blame alone.
But that isn’t the whole story. What’s driving students toward AI isn’t only fear of a bad decision, it’s a need for personal attention. As AI usage among our prospective students has climbed this year, our WhatsApp channel has filled up alongside it, not emptied out. Students want guidance tailored to them, and they’ll take it from whoever answers first and means it.
This is the psychology of the student as a buyer, shifting faster than most recruitment strategies can track. A generation raised on synthetic content doesn’t reward polish; it rewards proof and being heard. Peer testimony and a real person on the other end of a message now outperform a polished campus video. The brand isn’t the message. The evidence is.
The algorithm is watching geopolitics too
The third force is one the sector can’t optimise its way around: the shortlist a chatbot produces is shaped in real time by policy volatility. In the US, visa issuance to Indian and Chinese students fell roughly 60% and 46% this summer versus the 2017-2025 average, and NAFSA data shows new international enrolments falling 20% year-on-year this spring.
New September rules add a four-year completion cap and halve the post-study grace period. None of that is abstractpolicy to a model trained on current news, it’s a live input reshaping what “safe” and “worth it” mean.
Consultant Vincenzo Raimo calls it a “realignment” toward Europe, toward transnational education, toward markets that look stable rather than prestigious. An algorithm has no institutional loyalty; it reads volatility and adjusts the answer.
Multiple forces, one shift
Put together, this isn’t separate trends, it’s one. The student-as-buyer no longer trusts the seller’s own voice, so they’ve handed decision-making weight to a machine intermediary that at least sounds attentive. That machine, in turn, is reading a world of visa reversals and shifting alliances, updating its advice before a communications team drafts a statement.
The question this raises isn’t tactical. It’s existential. Institutions have spent years asking how to market more effectively to seventeen-year-olds. The sharper question is who, or what, is making the decision on their behalf, and whether anyone here has had a say in what it’s been told.
