Applicants are building their shortlist by asking an AI assistant, and most university course pages cannot be read by one. By the time somebody arrives on your site they have already decided you are worth looking at, and that decision happened somewhere your analytics cannot see.
No UK agency currently owns generative engine optimisation for higher education. Several hold traditional SEO pages for the sector. This page is about the other thing.
Test It Before You Read On
Three prompts. Two minutes. Substitute your own institution and subjects:
- “Which UK universities are best for [your strongest subject]?”
- “I have [realistic grade profile]. Which UK universities should I consider for [subject]?”
- “Tell me about studying [your flagship course] at [your institution].”
The first two tell you whether you are in the shortlist at all. The third is the one that unsettles people, because it shows the description of your course that a prospective student is being handed, assembled from league tables, forums and whatever your page happened to make extractable.
Run the third one on a competitor as well. The difference is usually instructive and it is usually not about quality of teaching.
Why Models Skip Course Pages
University course pages are among the worst-performing page types in AI search, and it is a structural problem rather than a content-quality one.
The information is in tabs. Modules, entry requirements, fees, placement details and career outcomes are behind accordions and JavaScript widgets. If it needs a click or a script to appear, assume a crawler will not see it.
The opening paragraph says nothing specific. Most course pages open with copy that would fit any course at any institution. There is no extractable answer, because the first sixty words contain no fact a model could quote.
Modules are a list of titles. “Advanced Research Methods” tells a model nothing. A sentence explaining what a student actually does in that module is the difference between a page that can be cited and one that cannot.
Entry requirements are a table with no prose version. The single most common question an applicant asks an assistant is some version of “can I get in with these grades”, and the answer is frequently unreadable on the page that holds it.
Clearing pages are built for a fortnight. They go up late, they are thin, and they are often excluded from indexing until results day, which is exactly the wrong sequence for a system that needs time to build confidence in a URL.
The fix is not more content. It is restructuring what already exists so an answer can be lifted from it, then marking it up so the structure is machine-readable.
League Tables, Wikipedia, And The Sources AI Actually Trusts
Here is the part universities find least comfortable and most useful.
AI answers about universities lean on third-party sources far more than on institutional marketing. League tables, Discover Uni, Wikipedia, UCAS, student forums and press coverage carry more weight than your prospectus, because they are independent and comparable across institutions. A model answering “which universities are good for engineering” has no reason to prefer your account of yourself over a ranking table.
Two things follow.
First, your own pages still matter, but for a different job. They are where the specific, checkable detail lives: entry requirements, module content, placement arrangements, accreditation, what the assessment actually is. A model reaching for that detail needs to be able to find and extract it. That is entirely within your control and it is where the fastest wins are.
Second, the corroboration layer is where the competitive positions are decided, and it is largely unmanaged in the sector. Wikipedia accuracy, Wikidata presence, the consistency of your institution’s description across sources, accreditation and professional body listings, academic staff profiles that establish genuine subject authority. Universities have more legitimate third-party authority than almost any other kind of organisation and are unusually bad at making it machine-readable.
That is Pillar 2 and Pillar 4 of the Synaptic Authority Engine, Knowledge Graph Presence and Third-Party Authority, and in this sector they matter more than anywhere else we work.
Working To The Admissions Calendar
Agencies that impose their own timeline on a university are a poor fit. The cycle is fixed and it is not negotiable.
Autumn to winter is when the work should be done: entity and extractability fixes across the courses that matter, while the applicant-facing pressure is lower.
Spring is when authority building has to be underway if it is going to be worth anything by summer, because corroboration accumulates over months rather than weeks.
Summer and clearing is a monitoring and rapid-response window, not a build window. Clearing pages need to exist and be crawlable well before results day, not on it.
The practical implication is uncomfortable and worth saying: a programme started in July will not change your clearing. It will change next year’s. If somebody tells you otherwise, ask them how third-party corroboration accumulates in four weeks.
Procurement, Frameworks And The Honest Part
What is straightforward. Pricing is published, so it can be specified and compared without a discovery call. Scope is defined with named deliverables. Engagements routinely involve a data processing agreement, a security questionnaire, least-privilege access to analytics and CMS, and working inside your change-control process.
What is not. MarGen is a small specialist firm. It is not currently on a higher education purchasing framework, and it does not hold ISO 27001 or SOC 2 certification. For some institutions that is an absolute gate, and it is better established in the first conversation than the fourth.
What that structure buys you is the other side: senior people on the work rather than on the pitch, and a supplier with no larger service line to grow into your budget from underneath.
The £2,950 Synaptic Audit is often the right procurement shape here, because it is a defined one-off below most institutions’ tender thresholds, it produces evidence you can take to a business case, and it is credited in full against a retainer if you continue.
Working With Your In-House Team
Universities have capable digital teams. They also have a CMS with genuine constraints, a governance structure, and a queue.
The division that works: MarGen owns the specialist layer (which changes matter for AI citation, in what order, and why, plus the entity and corroboration work), your team owns implementation on the platform they know. Recommendations are specified as actionable tasks rather than delivered as a report that somebody then has to translate.
Where a change is impossible in your CMS, we would rather know in week one and design around it than specify something that sits in a backlog for a year.
Pricing
| Step | Investment | Term |
|---|---|---|
| Free AI Visibility Audit | Free | None |
| Synaptic Audit | £2,950 one-off, credited against your first retainer month | None |
| Foundation | £2,950 a month | 3 months |
| Authority | £5,950 a month | 12 months |
| Dominance | From £12,950 a month | 12 months |
A single faculty or a defined group of courses usually sits at Foundation or Authority. An institution-wide programme across multiple faculties is Dominance-shaped, because it is multi-front by definition and each faculty has its own competitor set.
International Recruitment Is The Strongest Case
If one argument gets this funded, it is this one.
A domestic applicant has ambient knowledge of UK universities from school, family and league tables they have seen for years. An applicant in Lagos, Mumbai or Kuala Lumpur has none of that, and every reason to ask an assistant which UK universities are strong for their subject and would accept their qualifications.
That conversation happens entirely outside your marketing, frequently in a language other than English, and your visibility in it is almost certainly unmeasured. Non-English AI visibility is uncontested across the entire UK higher education sector, which makes it both the largest gap and the cheapest to open a lead in.
Proof
The closest structural analogue in our published work is a directory built across 37 UK cities, taken from zero to 66,000 impressions. The parallel is the shape of the problem rather than the sector: a large set of similar pages that each have to be individually extractable and individually corroborated, which is exactly the course-catalogue problem.
Method proof on our own site: zero to 295,000 impressions in 90 days.
We have not published a higher education case study, and we are not going to imply we have. If being the first reference client is a problem, that is a fair position and worth saying early.
GEO for Higher Education: Common Questions
Do applicants actually use AI to choose a university?
They use it to build the shortlist, which is the decision that matters. Very few people ask an assistant to pick a university outright. They ask which universities are strong for a subject, which offer a course with a placement year, which accept a particular grade profile, and what a specific course is actually like. By the time they reach your website they have already decided you are worth a look, and that decision was made somewhere you are not measuring.
Why do AI systems skip university course pages?
Because most course pages are built for a database, not a reader. The distinguishing information sits in tabs, accordions and JavaScript widgets, the modules are a list of titles with no explanation, entry requirements are in a table with no plain-English version, and the opening paragraph is marketing copy that would fit any course at any institution. There is often nothing extractable on the page, so a model answers from a league table or a student forum instead.
Which sources do AI systems trust about universities?
Mostly not yours. League tables, Discover Uni, Wikipedia, UCAS, student forums and press coverage carry more weight in AI answers about universities than institutional marketing does, because they are third-party and comparable across institutions. That is not a problem to be fought. It is the map of where the work has to happen, alongside making your own course pages genuinely extractable.
Can this be done in time for clearing?
Structural work on a defined set of courses can be done inside a term, and the entity and extractability fixes are the ones that move fastest. Citation authority takes three to six months, so a programme started in spring is working properly by clearing and a programme started in July is not. The realistic answer for a late start is to fix the highest-volume courses and the clearing pages themselves, then build authority for the following cycle.
How does this fit a university procurement process?
MarGen publishes its pricing, which makes it straightforward to specify and compare, and works to a defined scope with named deliverables. Engagements routinely involve a DPA, a security questionnaire and least-privilege access. Being plain about the limits: MarGen is a small specialist firm, is not currently on a sector purchasing framework, and does not hold ISO 27001 or SOC 2. If framework membership is a hard requirement, establish that in the first conversation.
Will you work with our in-house digital team?
That is the normal arrangement. Universities have capable web teams working inside a governance structure and a CMS that constrains what is possible. The specialist layer is knowing which changes matter for AI citation and in what order. The implementation usually belongs with the team who own the platform, and the output is specified so they can action it rather than handed over as a report.
How much does GEO cost for a university?
The ladder starts with a free AI Visibility Audit, then a £2,950 Synaptic Audit which is credited in full against your first retainer month. Retainers run from £2,950 a month on Foundation, £5,950 on Authority, and from £12,950 on Dominance. A single faculty or a defined group of courses usually sits at Foundation or Authority. An institution-wide programme across faculties is a Dominance-shaped engagement because it is multi-front by definition.
What about international recruitment?
It is the strongest argument for doing this work and the most neglected. An applicant in Lagos, Mumbai or Kuala Lumpur has no local knowledge of UK institutions and every reason to ask an assistant. They may also ask in a language other than English, where your visibility is almost certainly worse and almost certainly unmeasured. Non-English AI visibility is uncontested ground across the whole sector.
Start With Ten Courses
You do not need an institution-wide decision to find out whether this matters. Pick ten courses, the ones that recruit hardest, and we will show you exactly what an applicant is being told about each of them across all five platforms.
- Run the free AI Visibility Score on your domain first. One minute, no call.
- Book a conversation and bring the ten course URLs.
Related reading: enterprise GEO for multi-faculty programmes, how to measure AI search visibility for what to report to a board, why am I not in AI search results for the diagnostic, and GEO agency UK for the full service.