Wikipedia, LinkedIn, Press: The Sources LLMs Read About Your School

Why LLMs almost never read your website directly

ChatGPT, Perplexity and Gemini rarely form an opinion about your institution by crawling your homepage. They lean instead on a small set of third-party sources they have learned to treat as trustworthy — and a recent 2026 study puts real numbers on which ones.

A 2026 study by 5W Public Relations, reported via PR Newswire, found that Wikipedia and Reddit together drive over 25% of ChatGPT citations in the US. Notably, the Wall Street Journal, the New York Times and Bloomberg do not appear in the top 20 cited sources at all. This is a US dataset, but it is a useful proxy for how large language models behave across North America, including Canada: prestige alone does not earn a citation. Structural trust signals do.

Separate analysis from Profound's 2026 citation-pattern study sharpens the picture further. Wikipedia alone accounts for 7.8% of all ChatGPT citations and nearly half — 47.9% — of citations within ChatGPT's top 10 sources. Perplexity behaves differently, skewing toward LinkedIn, NIH and G2, and citing an average of 21.9 sources per response compared to ChatGPT's 10.4. The practice of optimizing for this behaviour has a name — Generative Engine Optimization, or GEO — and it is distinct from ranking on Google.

This matters for a specific reason: Schema.org markup, covered in our guide to structured data that makes your school visible in AI, governs what happens ON your own site. It tells an AI engine what your institution is. But it cannot tell the engine whether to trust what you say about yourself. That trust is built off-site, through sources the LLM did not have to take your word for. This article covers those off-site sources: Wikipedia, LinkedIn and Canadian press.

Wikipedia: the page you can't buy

A Wikipedia article is the single highest-weight citation source across ChatGPT, and no amount of budget or self-editing can force one into existence. It has to be earned through independent notability, and Canadian institutions need Canadian press coverage to get there.

Notability is not optional

Wikipedia's guideline for organizations requires "significant coverage in reliable secondary sources that are independent of the subject" — see the Wikipedia:Notability (organizations and companies) policy. A press release, your own website, or a paid placement does not count. What counts is a Maclean's feature, a Globe and Mail news story, or CBC coverage that discusses your institution in substance, written by someone with no financial relationship to you.

For a Canadian college or private university without a long public profile, this is the real bottleneck. You cannot skip to "have a Wikipedia page." You have to first generate the independent press coverage that makes a Wikipedia editor conclude the subject is notable.

Never self-edit — and here's why it backfires

Editing your own institution's Wikipedia page, or having a marketing staffer do it, is a conflict of interest that Wikipedia actively detects and flags. Wikipedia's Conflict of Interest guideline requires paid or affiliated editors to disclose the relationship, and undisclosed self-editing is routinely caught through account history, IP tracing and edit-pattern analysis. A flagged page often gets a visible COI banner — which is worse for AI trust signals than having no page at all, since it marks the source as compromised.

The correct route for a new page is the Articles for Creation (AfC) process: a draft submitted for independent editor review before it goes live, rather than a live edit to the encyclopedia. It is slower, but it produces a page that survives scrutiny.

• Commission or compile independent secondary coverage first — a Maclean's ranking mention, a Globe and Mail article, a CBC regional story
• Draft neutrally, in encyclopedic tone, with every claim tied to a citation — no marketing language, no "leading" or "renowned"
• Submit through Articles for Creation and let an independent volunteer editor review and publish it
• Disclose any paid or affiliated relationship explicitly on the talk page, as Wikipedia's policy requires

LinkedIn: the identity LLMs treat as verified

LinkedIn functions as a de facto identity layer for institutions in AI citation patterns, and it is currently the second most-cited domain across major AI platforms. According to ALM Corp's 2026 analysis of 325,000 prompts, LinkedIn ranks #2 among cited domains overall, and LinkedIn Articles specifically account for 50–66% of all cited LinkedIn content. Perplexity, per Profound's data above, over-indexes on LinkedIn relative to ChatGPT.

Company page completeness is the baseline

An incomplete LinkedIn company page — missing employee count, no program details, no recent posts — reads to an AI engine roughly the same way an unclaimed Google Business listing reads to a search engine: unverified. At minimum, keep the following current: official name matching your legal and Wikipedia name, website URL, employee headcount, industry classification, and a description that states accreditation status and program areas in plain language.

LinkedIn Articles from leadership carry outsized weight

A long-form LinkedIn Article published under your president's or provost's name is more likely to be cited than the same content published as a company page post. This is consistent with a broader pattern: personal profiles of leaders and faculty capture roughly 65% of organic reach on LinkedIn, compared to roughly 5% for company pages. A registrar or program director who regularly publishes about admissions trends, program outcomes or research is building a citation asset that a marketing department account cannot replicate on its own.

• Publish leadership Articles quarterly at minimum — enrolment trends, program launches, research outcomes, framed in the leader's own voice
• Encourage faculty and staff to maintain active personal profiles that mention their institutional affiliation explicitly and consistently
• Cross-link company page and personal profiles so LinkedIn's own graph — and, by extension, an LLM's crawl of it — can associate the two
• Keep post cadence steady rather than sporadic; LLM citation patterns favour sources with a consistent publishing signal, not one-off spikes

Press and media: the coverage that counts (and the coverage that doesn't)

Editorial coverage in an independent outlet is a citation-worthy source; a press release distributed by your own communications team is not, and LLMs increasingly distinguish between the two. The distinction is not subtle to a language model trained on the difference between a news article with a byline and a wire-service press release with a "contact us" footer.

Coverage that counts:

• A feature or ranking mention in Maclean's University Rankings — Canada's most cited annual post-secondary ranking
• A news story in The Globe and Mail covering a program launch, research result, or enrolment trend, written by a Globe reporter
• Regional CBC coverage of a campus event, research finding, or notable faculty appointment
• A citation or data mention in a Universities Canada report or member communication

Coverage that doesn't count, or counts far less:

• A press release published verbatim on a PR wire and picked up by aggregator sites with no editorial rewrite
• Sponsored content or "advertorial" pieces, even when they appear on a reputable outlet's domain
• Your own newsroom page, however well-written, since it fails the independence test that both Wikipedia notability and LLM citation weighting apply

The practical implication is that a communications strategy built entirely on press releases produces very little GEO value. Time and relationship-building spent pitching an actual Globe and Mail or Maclean's journalist a genuine story — a research first, a notable graduate outcome, a program that solves a documented labour-market gap — produces the independent, durable citation that both Wikipedia editors and AI engines are looking for.

Building your "source stack": a sequenced 3-step method

Wikipedia, LinkedIn and press are not parallel, independent tasks — they feed each other in a specific order, and starting in the wrong place wastes months. The sequence below reflects that dependency.

Step 1 — Earn press coverage first. Pitch a genuine story to a Maclean's or Globe and Mail reporter, or a regional CBC outlet, built around something concretely newsworthy: a research result, an accreditation, a graduate outcome tied to a labour-market trend. This step has no shortcut — it is the raw material every later step depends on.

Step 2 — Convert that coverage into durable, off-site assets. Use the resulting press citations to support a Wikipedia Articles for Creation submission, and cite the same coverage in LinkedIn Articles published under your leadership's names. This is also the point to make sure your on-site Schema.org markup references the same entities — press mentions, accreditations, program names — so on-site and off-site signals reinforce rather than contradict each other.

Step 3 — Measure and repeat on a schedule. Track whether the new sources actually move AI citation frequency using the KPIs described in our ChatGPT and Perplexity visibility KPI guide, and fold the source-stack work into the sequencing laid out in our 90-day reputation plan for higher education and the 90-day action plan to get cited by ChatGPT and Perplexity. For the underlying signals LLMs weigh when recommending an institution at all, see the LLM signals behind school recommendations.

Institutions with structured Schema.org markup on their own site achieve an average of +12 percentage points in AI visibility globally (Source: Skolbot GEO Monitoring, 500 queries x 6 countries x 3 AI engines, Feb 2026). That figure is a global finding, not Canada-specific, but it illustrates the same principle at work on-site that this article addresses off-site: structured, verifiable signals outperform unstructured claims, whether they live in your HTML or in a Wikipedia infobox.

FAQ

Why doesn't my institution's own website get cited by ChatGPT?

Your website is a primary source, and LLMs are trained to prefer independent, third-party verification over a subject's own claims about itself. Wikipedia, LinkedIn and independent press coverage all pass an independence test your homepage cannot pass by definition. Schema.org markup helps an engine understand your site; it does not substitute for external validation.

Can a private college get a Wikipedia page without national media coverage?

Yes, but it requires genuine independent secondary coverage, which does not have to be national. Regional CBC coverage, a Maclean's mention, or sustained coverage in a trade publication like University Affairs can satisfy Wikipedia's notability guideline for organizations. What will not work is a page built from the institution's own materials or a single press release.

Should marketing staff edit the institution's Wikipedia page directly?

No. Direct editing by affiliated staff is a conflict of interest that Wikipedia's policies are built to detect, and a flagged COI page carries a visible warning that undermines the very trust signal you are trying to build. Use the Articles for Creation process and disclose the institutional relationship on the talk page instead.

Does a LinkedIn company page matter if leadership already has personal profiles?

Both matter, but they are not interchangeable. Personal profiles of leaders and faculty capture roughly 65% of organic reach compared to roughly 5% for company pages, so leadership visibility drives most of the citation value — but a stale or incomplete company page still undermines AI trust in your basic identity data (name, size, programs).

How does this differ from what Schema.org markup does?

Schema.org markup structures what an AI engine finds ON your own site — your name, programs, accreditations, in machine-readable form. Wikipedia, LinkedIn and press build trust OFF your site, through independent parties who vouch for the same facts. You need both: markup without third-party validation is an unverified claim; third-party validation without markup leaves the AI engine unable to parse your own pages efficiently.

Wikipedia, LinkedIn and Canadian press coverage are not a public-relations afterthought — they are the trust infrastructure that decides whether ChatGPT and Perplexity mention your institution at all. None of the three can be bought outright, and all three take months to build properly, which is exactly why starting now matters.