AnswersFor an education provider

How do applicants find a course when they ask an AI assistant?

An applicant asks an assistant which universities are best for a subject, or which courses they could get into with the grades they expect. The answer names a handful of institutions and courses. What decides which ones are named is league tables, course databases and student forums, and whether your own course pages state the requirements and the deadlines clearly enough to be repeated.

What happens when an applicant asks an assistant where to study?

The assistant names a handful of institutions and courses for that subject, sometimes with a line about what it takes to get in, and stops there. An institution it does not name is off the shortlist before anybody has opened a prospectus.

This is a shortlist rather than a search. An applicant types a subject and a constraint, the best universities for something, or where the grades somebody expects would be enough, and the answer is a short list of named institutions with a sentence each.

Two jobs sit behind that one answer. Being named at all, which is decided by what the tables, the course databases and the student threads say about who teaches that subject. And being described accurately enough to survive the next question, which is decided mostly by what your own course pages publish.

Delphi counts a shortlist as an outcome here alongside the application itself, because the shortlist is the step an assistant genuinely produces. It is also the one an institution can be missing from without ever hearing about it.

Why do three different kinds of source all carry the weight here?

Because this is the one category where a ranking table, an education title and a student forum are all decisive at once. Most business types have two kinds of source that settle their answers. An education provider has three, and they are answering different questions about you.

The first are the ranking tables, the application services and the course databases. They hold the structured facts about who teaches what, on what terms, and what it takes to get in, and they are the sources an assistant reaches for when the question has a filter attached to it.

The second are the league tables and the education press, which are cited heavily on the plain question of which institution is best for a subject. That is the question that decides the shortlist, and it is answered somewhere no institution controls.

The third are the applicant forums and the student threads, where institutions are argued over by name. That is where the reputation an assistant repeats is actually formed, and it is the source that most often carries a sentence about what a place is like to study at rather than a figure about it.

They disagree with each other, and that is the useful part. A table ranking on one measure, a database listing a requirement, and a thread about what the teaching is really like are three answers to one question, and an assistant blends them into a paragraph. An institution strong in one and absent from the other two gets a thinner answer than its position deserves.

Do entry requirement questions count if no institution is named?

They count for perception rather than for recommendation. Entry requirement, funding and process questions are where an applicant forms a view long before a shortlist exists, and they are measured apart from the questions where an institution is being chosen.

Delphi treats four families of question as core for an institution, and this is the one it holds out of the score for being chosen. An applicant asks what a subject requires, whether funding exists, when the deadline falls and how the application itself works, and most of those answers name no institution at all.

Those answers are measured for what they say and whose wording they use, and they are held apart from the score for being chosen, because a question that names nobody has no recommendation in it to win.

Which makes the requirements and funding pages worth writing well for their own sake. They are read at the point where the decision starts, and they are the reason an assistant can state what your course asks for in your own wording rather than in a paraphrase picked up somewhere else. The same subjects come back in the constraint questions where an applicant does expect a list of institutions, and those are scored.

What is checked once an assistant has named your course?

Whether it named the right course, stated the entry requirements and the deadlines the way you publish them, and offered a route that resolves. Delphi calls this last step Applied, and it reads the route to an application rather than a checkout.

The last of the five steps means something different in every industry, and for an institution it is called Applied. It reads the handover: does the answer point at a course you actually run, does it repeat your requirements and your deadlines, and does the link it offers go somewhere real.

A wrong requirement is the expensive version. An applicant told they need a grade they will not have takes themselves off the list, and one told they need less than you ask arrives at a form they cannot complete. Both happen quietly, and neither leaves a trace anybody at the institution would see.

So the failure this step is looking for is being recommended and then leaving somebody unable to find what it takes to get in. Being named is the first job. Being followable is the second, and it is the one that turns a shortlist into an application.

Does it matter where the campus is?

Yes, and it is one constraint among several rather than the whole question. An institution is bound to premises, so location sits alongside grades, fees, course length and accreditation in the list of things an applicant types.

For a business somebody looks up by where it is, place is the whole question. An institution is asked for by subject, and place arrives as one filter among the five an applicant types, which changes what has to be true for you to survive it.

What that changes is the shape of the work. An assistant has to be able to answer all five about you at once to keep you on the list, and failing any one of them removes you as completely as never being named.

It also means the where is worth stating plainly, and stating per course. An institution teaching the same subject on more than one site needs an assistant to be able to say which one, because an applicant filtering on location is filtering on the answer to that.

What is read on the institution website itself?

Course pages, and whether they resolve and show they are current. They are the only authoritative statement of what a course requires and when it closes, so they are what an assistant reads when the answer has to be exact.

Two readings carry it. Whether the course, requirement and application links resolve, and whether the course pages show they are current, which is what decides whether the requirements stated on them can be trusted at all.

Freshness carries more weight here than in most categories, and that second reading is why. A course page that cannot show it is current is not empty. It is confidently wrong, and confidently wrong is the version that gets repeated.

What gets read is the material an applicant actually needs: the course, what it takes to get in, when it closes, and where to apply. That is the set an institution is measured on, because a course is a thing to apply for rather than a listing to buy, and a place on one is not stock.

Published fees are real and they matter to every applicant, and they usually sit in a form an automated check cannot confirm. The same is true of the course data an application service consumes. Delphi reads both as unmeasured rather than as gaps in your site.