This week
In 2025, downloading a brochure overtook talking to an advisor as candidates' preferred first contact with a business school, in a dataset of 102,900 verified leads, and the leads in that dataset arriving from ChatGPT convert at nearly four times the rate of those from Facebook (source). And on Thursday 20 August IDP Education, the largest listed company in international student placement, reported full year placements down 27% and said it is planning for volumes across its six key destination markets to fall by another 20% to 30% in the year to June 2027 (source). Candidates are finding their own way in; the paid human network that used to guide them is shrinking. The Spotlight puts the two halves together.
Clearing is still running. By Monday 24 August, 60,980 international students had been accepted, against 59,110 at the same point in 2025 (source). The numbers below show where that growth is coming from; it is the same place as on results day.
The numbers
IDP placed 54,219 students in the year to June 2026, down 27%, and is planning for volumes across its six key destination markets to fall by another 20% to 30% in the year to June 2027. (source) The accounts published on Thursday 20 August show revenue of AUD 795.4m, down 11%, income from placing students of AUD 362.8m, down 17%, and profit of AUD 13.3m, down 74% after AUD 36m of restructuring and transformation costs (source). Of the four big destinations, Australia held up best, down 9%; the UK was next, down 20%; the US fell 67% and Canada 80%. The two small markets, Ireland and New Zealand, grew by 2% and 1%. The company's planning assumption is no further change to visa policy settings (source). Takeaway: IDP's guidance covers all six of its destination markets together; its own placements are now concentrated in Australia and the UK. Test your agent-channel forecast for 2027 against the 20% to 30%, and ask partners for a UK-specific view rather than a global one.
59% of prospective international students say the UK is expensive; 10% say it is affordable. (source) QS's International Student Survey, fielded from 5 January to 6 April 2026 with responses from over 42,115 students in 191 locations, still puts the UK's core draw first: 58% cite high quality teaching. But the fastest rising decision factors since 2022 are the ability to get a visa after graduating, the ability to get a study visa, and good graduate employment options. Reputation still attracts interest, but visas and value increasingly decide whether it becomes an application. Takeaway: your prospectus sells teaching quality; the decision is increasingly made on visa confidence and value for money. Check your ten most visited web pages: how many answer a question about cost, visas or jobs before the reader has to scroll?
60,980 international students had been accepted through UCAS by Monday 24 August, up 1,870 on the same day in 2025. (source) Acceptances from China are up 1,920 over the same period, to 16,830, so every other country combined is still slightly down, by around 50. Issue 17 measured this concentration on results day, when China held 25.8% of the international intake; eleven days of Clearing later, the share is 27.6%. UCAS rounds its counts to the nearest 10, so treat small differences like that 50 as approximate.
Policy watch
Eight US organisations sued on 18 August to stop the end of duration of status. (source) The coalition, led by NAFSA and the Presidents' Alliance, filed in the District of Massachusetts, with a motion for a preliminary injunction before the rule, covered in Issue 17, takes effect on 15 September. If the injunction is granted, part of the projected shortfall of up to 110,000 US students in autumn 2026 could unwind; if not, the diversion window Issue 17 described stays open into January 2027.
Taiwan's international enrolment rose 17% in 2025/26, to 140,420, with a government target of 320,000 by 2030. (source) Regional systems are now competing for the same Asian students as the big four destinations.
The signal
IDP's results are one company's, but the company is the market leader. Placements down 27%, placement income down 17%, roughly 1,250 roles removed, and a plan built on volumes falling further (source). Some of that fall is the market shrinking; some may be share lost to the direct and self-service routes this issue describes. Either reading is bad news for the traditional funnel, and whatever IDP is having to do, its smaller competitors are doing with less financial room. Issue 17 reported Crizac, a listed Indian aggregator, paying 1.7 times revenue for UK linked distribution (source). Contraction and consolidation point to the same conclusion: the layer of intermediaries between student and university is getting smaller and more concentrated.
There is a twist in QS's UK survey: students say agents and consultants are becoming more important to them, even as the firms supplying that guidance shrink (source). Both things can be true. Demand for guidance has not fallen; it will just be met by fewer, larger companies. And a student can want an agent's help while still preferring to make first contact with a school on their own, which is what the House of Education data shows. That is the other half of the squeeze, and it is where the Spotlight goes next: the paid human channel is contracting from above while candidates remove the human from the start of their own journey. A 2027 recruitment plan built on the 2019 funnel will find both ends have moved.
Spotlight · Sector · The candidate arrives pre-qualified
In 2025, downloading a brochure overtook talking to an advisor as candidates' preferred first contact with a business school. Leads referred by ChatGPT convert at nearly four times the rate of the network's benchmark paid channel, Facebook. The report's reading: these candidates arrive with the comparison already made.
The data comes from House of Education, a Stockholm marketing firm whose Digital Advisor questionnaire sits on business school websites across Europe. Its Candidate Insights Report 2026 covers 102,900 MBA and EMBA leads collected in 2024 and 2025 from more than 200 countries, every one with a verified email and phone number, plus a first cut of Q1 2026 (source). House of Education sells the tool the data comes from, which is true of most vendor research; what earns this report a Spotlight is what it measures. This is behaviour recorded at the moment of engagement, not a survey of students who already enrolled, and we have not seen published figures on AI-referred recruitment anywhere else.
Finding one: fewer candidates want to start with a person. In 2025, downloading a brochure overtook talking to an advisor as the preferred way to hear more from a school, globally. In the Q1 2026 cut of 11,780 candidates, 29% chose the brochure, 21% an advisor, 19% the website, and 14% each an online event or nothing at all. Year on year, the share of candidates choosing an advisor fell by around five percentage points in Asia, three in Africa and half a percentage point in Europe; only the Americas rose, by around one percentage point.
Source: House of Education, Candidate Insights Report 2026 (MBA and EMBA), Q1 2026 channel preference distribution.
Finding two: the leads referred by ChatGPT behave differently. House of Education tracks where its verified leads come from. Referrals carrying a ChatGPT tag went from a standing start in late 2024 to roughly 180 a month by January 2026, with the sharpest jump in autumn 2025. The report links the growth to ChatGPT adding sources and links to its answers, a feature that arrived in late 2024; with only one year of data, the jump may also partly be the start of the autumn research season. Whatever drives the volume, the leads convert differently. Against Facebook, the network's benchmark paid channel, ChatGPT-referred leads convert at 29.17% against 7.77%, and spend 5 minutes 38 seconds on task against 3 minutes 8 seconds. The report's reading is that these candidates arrive having already researched, compared their options and refined their decision, effectively pre-qualified by their AI-assisted exploration.
Source: House of Education, Candidate Insights Report 2026 (MBA and EMBA).
The absolute numbers are small: roughly 180 tagged referrals a month across an entire partner network, and tagging undercounts AI influence rather than overcounting it, because a student who reads a ChatGPT answer and then searches for the school directly leaves no tag. The sample is one-sided by construction: every candidate in it arrived through a website questionnaire, so people who begin with an agent or at a fair never enter the data. Conversion here means completing House of Education's own sign-up questions, not applying or enrolling. Part of the conversion gap is also what any organic channel would show against paid social, since people who find their own way to a form were already interested; the report offers no organic search comparison that would separate the AI effect from that. And the setting is European business schools, with a strong French weighting, recruiting MBA candidates, not UK undergraduate admissions. These are early numbers from House of Education's network alone. They show something starting to happen; they do not yet show how big it is.
The UK's biggest student survey points the same way. QS's International Student Survey 2026, fielded between 5 January and 6 April with over 42,115 respondents, finds discussions with admissions staff declining as an information source for UK bound students, social media rising more than any other source since 2022, over 50% of UK bound students using Instagram for research, and printed prospectuses falling from 20% in 2022 to 17% in 2026 (source). The two reports appear to disagree about brochures; they do not. QS measures which sources students use in general, where the printed prospectus is fading. House of Education measures the preferred first contact with one chosen school, where a downloaded brochure now leads. Both describe students serving themselves. QS adds one more pressure: the share of students expecting to know the outcome of their application within 24 hours has risen from 11% in 2022 to 19% in 2026.
And the behaviour does not begin at application. HEPI's third Student Generative AI Survey, published in March 2026 from fieldwork with 1,054 UK undergraduates, found 95% now use AI in at least one way, up from 66% in 2024 and 92% in 2025, and 94% use generative AI to help with assessed work (source). The applicant who researches universities through ChatGPT is simply the student HEPI describes, a year earlier in the journey.
A search engine sent students to the university's website, and the website did the persuading, where the university could see it. An AI assistant does the persuading before the click. The candidate arrives with the comparison already made, which is exactly what the conversion gap in this data shows: fewer arrivals, far more of them ready. Issue 8 measured the sector's conversion problem from intent data; this is evidence of a channel that sidesteps it, because the qualifying happens before the university spends anything at all.
Nothing here needs a vendor to act on, and that is the point for institutions. Every university can run this measurement on itself for free: referrals from ChatGPT, Gemini, Perplexity and Copilot are already sitting in your web analytics, and their enquiry to application conversion can be set against your paid and organic channels in an afternoon. If a lead that costs nothing converts at a multiple of a lead you paid for, that belongs at the top of the 2027 marketing budget conversation.
Things to think about this week
Pull twelve months of referral data and group traffic arriving from ChatGPT, Gemini, Perplexity and Copilot into a single named AI channel. Measure volume by month and enquiry to application conversion against your paid social channel and your organic search traffic; the organic comparison is the one that separates the AI effect from ordinary self-selection.
Ask ChatGPT and Gemini the questions your applicants ask: fees, entry requirements, scholarships, visa sponsorship for your three biggest courses. Check what they cite. If the answer quotes a competitor's page or an out of date fee, that is a content fix with measurable recruitment value.
Check your time to first personal response against the 24 hour expectation in the QS data. A pre-qualified enquirer is further down the funnel; making them wait three days wastes the advantage the channel gave you.
Test your agent channel for 2027 with IDP's guidance: the market leader plans for volumes across its markets 20% to 30% lower. If an agent partner's forecast assumes growth, ask for the UK-specific case for it before it goes in your plan.
Jobs · Who's hiring
The last word
The sector has spent a decade professionalising the human funnel: agent networks, advisor teams, fairs, response scripts. On Thursday 20 August the largest company in that funnel planned for the market it serves to shrink by up to 30%, and new behavioural data on AI-referred recruitment showed candidates building their shortlists in conversations no fair reaches and no dashboard fully sees.
The uncomfortable part is that this channel cannot be bought. There is no way to pay to be the answer ChatGPT gives. There is only being accurate, specific and worth citing where the machines read, and fast where the human finally appears. Both are unglamorous operational work, and both can be measured with the analytics already in place. The universities that measure it first will plan 2027 knowing something their competitors do not.
If this was forwarded to you, subscribe to receive every Tuesday at 7am UK time.
Sources
