Why AI engines ignore most university websites
ChatGPT, Perplexity and Google AI Overviews do not rank web pages. They synthesise answers from vast corpora and cite sources they deem reliable, structured and factually verifiable. Most university content still fails on all three counts.
In Canada, only 29% of ChatGPT responses about higher education name a specific university (Source: Skolbot GEO Monitoring, 500 queries x 6 countries x 3 AI engines, Feb 2026). On Perplexity, that figure rises to 38%, especially for institutions with strong profiles in OUAC, Universities Canada and the Maclean's university rankings. The remaining responses are generic summaries about "top universities in Canada" with no institution named. Your content exists, but AI engines cannot extract anything citable from it.
Four factors separate citable content from invisible content: technical structure, data specificity, source authority and answer clarity. Each one is within your institution's control, whether you recruit nationally or compete regionally in Ontario, British Columbia, Alberta or Quebec.
What makes content "citable" by an LLM
Structure beats length every time
An LLM does not read a page from top to bottom. It extracts answer fragments from recognisable patterns: question-answer pairs, comparison tables, definitions framed by semantic markup. A long page with no explicit structure is less likely to be cited than a concise page with descriptive H2s, a data table and a marked-up FAQ.
Structural signals that LLMs exploit:
| Signal | Impact on citability | Implementation difficulty |
|---|---|---|
| FAQ marked up in JSON-LD | High: direct extraction | Low |
| Tables with descriptive headers | High: comparable data | Low |
| H2/H3 phrased as questions | Medium: semantic matching | Low |
| Schema.org EducationalOrganization | High: entity identification | Medium |
| Sourced numerical data | High: verifiable facts | Medium |
Specificity wins over superlatives
Content claiming "our university offers exceptional student support" will not be cited. Content stating "93% of 2025 graduates found employment within 6 months, median salary CAD 57,000, co-op option available, 312 survey respondents" gives AI engines evidence they can reuse.
Data points AI engines actively look for on Canadian university websites:
- Employment and graduate outcomes, with methodology and sample size
- Tuition and ancillary fees by province, residency status or programme
- Official accreditations, designations and institutional memberships
- Rankings with source and year (Maclean's, QS, THE)
- Student numbers, co-op availability, campus location, application channel and intake timing
4 techniques to make your content citable
1. Implement Schema.org on key pages
Universities with structured Schema.org markup achieve an average of +12 visibility points in AI engine responses (Source: Skolbot GEO Monitoring, 500 queries x 6 countries x 3 AI engines, Feb 2026). The EducationalOrganization markup turns your institution into an identifiable entity. The Course schema does the same for each undergraduate, graduate or continuing education programme.
For the full technical implementation guide, see our Schema.org guide for universities.
The minimum implementation covers three schemas:
- EducationalOrganization on your homepage and About page
- Course on each programme page
- FAQPage on FAQ pages and blog articles containing Q&A sections
The fields that matter most to LLMs: accreditation, numberOfStudents, aggregateRating, alumni and programPrerequisites. These are the data points ChatGPT cross-references with the Office of the Privacy Commissioner of Canada, the official text of PIPEDA, provincial application centres such as OUAC, and public ranking data to validate reliability.
2. Structure every page with direct answers
AI engines operate on a question-answer model. To maximise citation probability, each H2 should pose or imply a question, and the first 1-2 sentences must answer it directly. The rest of the paragraph adds context and nuance.
Before:
"Our School of Management combines academic rigour with experiential learning and strong employer links."
After:
"The Bachelor of Commerce at [University] is a 4-year degree costing CAD 9,450 per year for Ontario students, with a 93% employment rate at 6 months and co-op availability in years 3 and 4 (2025 outcomes survey, 312 respondents). Applications are processed through OUAC, and the programme publishes both admission averages and internship participation rates."
The second version contains six verifiable data points. The first contains none.
3. Build comparison tables with your data
Tables are the most extractable format for an LLM. A clean table with clear headers and numerical data will be preferred over a narrative paragraph containing the same information.
Example of a citable table for a programme page:
| Criterion | Bachelor of Commerce | MBA |
|---|---|---|
| Duration | 4 years | 20 months |
| Annual tuition | CAD 9,450 | CAD 41,000 |
| Employment rate at 6 months | 93% | 95% |
| Median starting salary | CAD 57,000 | CAD 86,000 |
| Co-op / work-integrated learning | Yes | Internship option |
| Intake size | 260 | 70 |
Publish these tables on your programme pages, not just in downloadable calendars or PDF viewbooks. AI engines do not reliably parse gated documents.
4. Add marked-up FAQ sections
An FAQ section serves two purposes: it answers the questions prospects ask AI engines, and FAQPage JSON-LD markup enables structured extraction.
The common mistake is writing brand FAQs ("Why choose our university?") instead of informational FAQs ("What admission average is competitive?", "Does the programme include co-op?", "What are the tuition fees for out-of-province students?"). AI engines favour the latter.
To diagnose your current situation, use our ChatGPT visibility diagnostic tool.
How to measure whether your content is being cited
Checking whether AI engines cite your university requires a systematic approach.
3-step testing protocol
-
Identify your 20 strategic queries: the questions prospects ask about your university, programmes, city and sector. Examples: "best business school in Ontario", "top co-op university in Canada", "tuition [university] 2026".
-
Test across 3 AI engines: submit each query to ChatGPT, Perplexity and Gemini. Record whether your university is mentioned, whether the information is accurate, and whether sources are cited.
-
Track monthly evolution: LLM corpora evolve. Content published or updated today may take 4-8 weeks to be integrated. Measure monthly to identify trends.
Key metrics to track
| Metric | Target | Measurement frequency |
|---|---|---|
| Mention rate (brand queries) | >80% | Monthly |
| Mention rate (generic queries) | >20% | Monthly |
| Accuracy of cited information | 100% | Monthly |
| Sources cited (Perplexity) | >2 pages from your site | Monthly |
For a complete methodology on tracking your AI visibility, see our GEO guide for universities.
Before and after: optimising a programme page
An Ontario university wanted ChatGPT to mention its BCom programme when prospects asked "best undergraduate business programme with co-op in Canada".
Before optimisation:
- Programme page without Schema.org
- Narrative text with no numerical data
- No FAQ section
- No comparison table
Result: ChatGPT never mentioned the university for this query.
After optimisation:
Coursemarkup witheducationalLevel,provider,accreditation- Table with tuition, duration, co-op participation, employment rate and median salary
- Marked-up FAQ with 5 questions (OUAC code, admission average, co-op, scholarships, tuition)
- Link to OUAC and Maclean's as authoritative comparison sources
Result at 8 weeks: ChatGPT cites the university in 3 out of 5 responses for the same query. Perplexity links to the programme page as a source in 4 out of 5 cases.
This correlation between structured markup and citability holds across our full panel. For the technical mechanisms, our article on structured data for universities details each schema.
FAQ
How do I check if ChatGPT already cites my university?
Test 20 strategic queries directly in ChatGPT. Record every mention of your institution, the accuracy of the data and the presence of links. Repeat monthly to track changes. Perplexity is simpler to audit because it displays its sources beneath each response.
How long before optimised content gets cited?
Between 4 and 8 weeks after publication or modification. Perplexity usually reacts faster because it queries the live web. But the page still needs structured data, current tuition tables and credible external references.
Is Schema.org markup enough to get cited?
No, but it is necessary. Markup identifies your university as a verifiable entity. Without it, AI engines must extract this information from raw text, with a high error rate. Markup alone does not replace specific, data-rich, well-structured content.
Should I optimise for ChatGPT or Perplexity first?
Both, as the techniques overlap. But if you must prioritise, start with Perplexity: it cites sources explicitly, making tracking straightforward. Optimisations that work for Perplexity also benefit ChatGPT.
Which pages on my site should I optimise first?
Your homepage, the three most-enquired programme pages, your admissions page and your FAQ page. In Canada, the highest-priority pages are usually the ones that explain application channels, tuition by student category, co-op and employability outcomes.
Is your university cited by ChatGPT? Test your AI visibility for free


