Ask ChatGPT or Perplexity a shortlist question about programs in your region. Most likely your institution is not in the answer, and if it is, the citation points at a rankings site or a Reddit thread rather than your own site. The gap is not about brand strength or domain authority. It is structural. AI engines cite content they can extract, verify, and attribute, and most .edu sites are built for humans browsing, not for engines extracting.
Here are the five structural reasons universities lose the citation game, and the specific fix for each.
1. Program pages are prose first, question shaped never
The typical program page opens with a paragraph of aspirational copy, a photo carousel, and a "Learn more" button. The information a prospective student wants — outcomes, admissions requirements, cost, curriculum — is buried three scrolls down as flowing prose. Engines cannot extract a citable answer to "what is the acceptance rate for X program" from that structure.
The fix. Restructure priority program pages around question-shaped H2s with direct answers in the first sentence beneath them. "What does the program cost." "What are the admissions requirements." "What outcomes do graduates see." Keep the prose, but front-load the answer. Do not redesign, do not replatform. This is a content-model change your CMS already supports.
2. Critical information is trapped inside PDFs
Program brochures, financial aid tables, admissions handbooks, and academic catalogs all live in PDFs on most .edu sites. AI engines technically read PDFs, but they read them worst, extract from them least, and rarely cite them. So the numbers your recruiters wish were quoted — tuition, aid, class size, faculty ratios — never appear in AI answers.
The fix. Liberate the top ten most-linked PDFs into HTML pages. Keep the PDF for print or accessibility, but publish the same information as a real web page with headings, tables, and schema. Any information you would want an admissions counselor to be able to quote out loud should exist as HTML on your site.
3. There is no FAQ, Course, or Organization schema
Structured data is how the modern web tells engines "this block of text answers this question, this course belongs to this school, this program has this duration and this credential." Without it, engines are guessing. Most .edu sites either ship no JSON-LD at all or ship a generic Organization block on the homepage and stop.
The fix. Add three schema layers in order of impact. FAQPage on every question-shaped page. Course on every program page. BreadcrumbList sitewide. This alone typically moves an institution from unquotable to quotable inside a quarter.
4. No llms.txt or machine-readable layer
llms.txt and llms-full.txt are the emerging convention for telling AI crawlers what content on your site matters, in what order, and how to interpret it. They are cheap to ship, they are already respected by the major engines, and almost no .edu site has them.
The fix. Publish /llms.txt at your root domain with a prioritized index of your top program pages, key admissions pages, and canonical about pages. Add /llms-full.txt with the full text of those pages inline. This is a one-week project for most institutions and pays out for years.
5. Content is fragmented across subdomains and templates
Universities routinely have one program described across four templates: the catalog entry, the marketing site, the department subdomain, and the graduate school subdomain. Each version says slightly different things about tuition, deadlines, and outcomes. Engines see the disagreement and either pick the wrong version to cite or skip your institution entirely.
The fix. Pick one canonical page per program, link every other version at it with a rel="canonical" tag, and align the numbers. You do not need to consolidate subdomains overnight. You do need every version to point at one source of truth.
The order to fix these in
If you can do only one thing, do schema and FAQ restructuring on your top ten program pages first. That is the highest-leverage move for the lowest technical spend. Ship llms.txt second. PDF liberation third. Content deduplication fourth. Subdomain consolidation is a multi-quarter project — do it deliberately when governance allows, not as a reaction to AI search.
If you want the full playbook, read our practical AEO guide for higher education or scope DoodleWeb's higher-ed AI search visibility service. If you are still calibrating on how AI actually fits into the funnel, how students use AI in their college search is the piece to read first.
Seattle, WA
A full-service digital agency working in WordPress, Drupal, Shopify, Webflow, React, and React Native. We partner with universities, governments, and growing brands to ship sites and products that hold up after launch.
