A student chooses a programme by answering two questions in order: is this recognised, and what does it lead to. Fees, campus, curriculum and faculty are all weighed against those two answers rather than judged on their own. AI assistants have converged on the same order, because they read the same public evidence. The institution whose recognition status and outcome data sit in plain readable text gets named in the answer. The one whose facts live inside a prospectus PDF does not.
Two questions, in that order
A student chooses a programme by answering two questions in order: is this recognised, and what does it lead to. Fees, campus, curriculum and faculty are all weighed against those two answers rather than judged on their own. AI assistants have converged on the same order, because they read the same public evidence. The institution whose recognition status and outcome data sit in plain readable text gets named in the answer. The one whose facts live inside a prospectus PDF does not.
Scale is what put those two questions first. The British Council's 2026 summary of the Ministry of Education's All India Survey on Higher Education reported total enrolment of 45 million and a gross enrolment ratio of 30 per cent. A system that large stops being a place where a student applies to the institution nearby and becomes a category the student researches before choosing.
That changes who does the shortlisting. When there is one obvious option, the buyer researches your institution. When there are thousands, the buyer researches the category, arrives at your name already holding a shortlist, and applies tests you did not write.
What are students and parents actually comparing?
Six things, and they are not the six most prospectuses lead with. Accreditation and recognition. The placement record and the basis it is calculated on. Fees, funding and scholarships. Curriculum and faculty. Admission criteria and deadlines. Alumni outcomes.
Notice how they divide. Recognition and outcomes are usually the parent's questions, and curriculum, faculty and campus life are usually the student's. A page written for one reader answers half a household, which is why so many enquiry conversations stall on a question the website already contained but did not surface.
Notice, too, that cost never gets judged on its own. It is judged against outcome. A fee with no outcome attached settles nothing, and an outcome with no cohort attached settles nothing either. The pair is the unit of comparison, and publishing one half is close to publishing neither.
Recognition is the question that hedges the answer
Careers360 reported in October 2025 that 11,268 colleges had completed a first cycle of NAAC accreditation, against 45,473 colleges in the country on the All India Survey on Higher Education data it cites. Whatever else that gap means, it means recognition status is a live question a student cannot assume the answer to, and knows it.
An assistant reading your programme page for recognition has three outcomes available. It finds a plain statement and repeats it. It finds a registry entry and reconciles the two. Or it finds neither and hedges. The third is the expensive one, because a hedge does not read to a parent as missing data. It reads as a warning, and a warning about recognition ends the consideration rather than delaying it.
The fix is more specific than adding a badge. Recognition is granted per programme, per body and per period, and it expires. An accreditation logo in a footer is not a checkable statement about the course in front of the student. The accrediting body, the entity or programme accredited, the grade or status, and the validity period, stated as text on the programme page, is.
Why does a placement figure with no denominator get dropped?
Because a percentage on its own does not describe anything. A placement rate is a different number depending on whether it counts the eligible cohort or the whole class, offers or acceptances, what share of graduates responded to the survey, and which period it covers. Two institutions can publish the identical figure and mean genuinely different things, so a student cannot compare them and neither institution can defend its own.
So the requirement is not to hold the data. It is to publish the basis next to the number: class size, response rate, offers or acceptances, and the period. That is the difference between a claim an engine hedges and a citation it repeats with your name on it.
The same failure repeats on fees. A headline figure with no instalment schedule, no scholarship criteria and no statement of what is excluded, hostel, examination, materials, is a number a family cannot plan against. And it repeats on rankings, where an old position stays crawlable and gets quoted long after the table has moved. Both are cases of a fact separated from the qualifier that makes it usable.
The research now runs through an assistant
ICEF Monitor reported in October 2025 that in IDP Education's Emerging Futures survey of 7,900 current and prospective students, 54% plan to use AI to decide which institution to study at, up from 35% who planned to use AI for the same decision a year earlier. A behaviour that moves that far in twelve months is not a trend to watch. It is a channel that already exists.
It has not replaced your website. Ruffalo Noel Levitz reported in 2025 that 91% of students use campus websites to research institutions, and that 1 in 4 students say it is hard to find information about specific majors. The assistant is not competing with your programme pages. It is reading them, and that second figure is the constraint it inherits.
It has not replaced people either, and honest planning should say so. The same ICEF Monitor report cited EAB's 2025 survey of more than 20,000 US high-school students, in which 3% named AI chatbots as their most trusted source, against 30% for online resources such as college websites and search data.
Read the three together and the shape is clear. The assistant runs the first pass, your website supplies the evidence it reads, and a counsellor or a campus visit closes. The exposure sits entirely in the first pass, because failing there produces no enquiry, no name and nothing in the funnel for anyone to review.
What belongs on a programme page
Five blocks, all of them as text on the page rather than inside a download. Curriculum and faculty: what is taught, by whom, with what qualifications. Fees and funding: total cost, instalments, scholarships, and what is excluded. Accreditation, stated plainly. Outcomes with a cohort attached. And an application path.
The prospectus is the block most institutions get wrong, because a PDF looks like publishing. It is fetchable in principle and loses in practice. A designed fee table arrives as layout rather than as data, it updates on an annual cycle, and the version an engine is holding is frequently last year's. The same facts in an HTML table win, and the institution that lost did not know it was in a contest.
The application path is the block most institutions forget. An assistant that has run the comparison and decided in your favour still has to hand the student somewhere. Criteria, deadlines, and a genuine first step. If the answer can only end with contact the university, the work you did upstream converts for whoever published a form.
How do you tell whether any of this is working?
Track the ladder, not a score. For any question an applicant asks, your position is one of five states: the answer lists institutions and yours is absent, your institution is named with no programme attached, your programme page is cited for a fee or a criterion, your programme is in the comparison the student reads, or your programme is recommended for this student's goal. Those are five different problems with five different fixes, and averaging them into one number hides which one you have.
Run the questions applicants actually type, repeatedly. Placement record and average salary. Is this degree recognised by employers. Online degree versus on campus degree for working professionals. Is a masters degree worth the cost. Best universities for undergraduate admission. Each of those fans out into the six comparisons above before the engine writes a word.
Record two things per run: where you sit on the ladder, and which sources the answer cited. The second is the actionable half. If a ranking table, an accreditation registry and a student forum are doing the work, then the next month of Answer Engine Optimization belongs on those three surfaces and not on a homepage refresh.
One honest limit. An engine's answer moves between runs, so a single read is an anecdote rather than a measurement. Repetition across weeks, against the same prompts, is what turns it into evidence you can act on.