What ChatGPT Tells a 17-Year-Old About Your University

By Midya U · Midya U Advisory

Somewhere right now, a 17-year-old is deciding where to apply. She doesn't start at your website. She opens her phone and asks ChatGPT whether your university is good for the program she wants, whether the city is safe, what it all costs, and whether graduates actually get jobs. The answer comes back in five seconds.

You have never seen that answer.

I run this exercise with marketing and recruitment teams, and the same two problems show up every time.

Problem one: wrong answers

AI assistants build their picture of your institution from whatever they can read — Reddit threads, agent websites, old news coverage, directory listings, ranking sites, and your own pages. When those sources disagree, the model picks something. I've watched these tools quote tuition figures that are years out of date, recommend programs that were suspended in the last round of cuts, and describe a campus that hasn't looked like that in a decade. The student reading it has no way to know. Most won't check.

Problem two: silence

Ask an AI assistant to name good universities in your province for a specific field, and see whether you show up at all. When a model has little to go on, it defaults to the biggest names and the most-written-about schools. You can lose an applicant before you ever knew you were being considered. No declined offer, nothing your CRM will ever log. The funnel didn't leak; it never started.

The check I recommend — it costs nothing

1. Write down the ten questions your applicants actually ask. Cost, safety, co-op, housing, visas, "is it worth it." Pull them from your recruitment and admissions inbox, they're all there.
2. Ask ChatGPT, Gemini, and Claude each of them, in fresh sessions. Fresh matters — you want the answer a stranger gets, shaped by nothing.
3. Compare against reality. Mark every wrong fact and every question where a competitor was named and you weren't.
4. Ask the tool where the information came from. The sources it names are your repair list.

An hour of this tells you more about your discoverability in 2026 than most analytics dashboards will.

What to fix first

Make your program pages state facts plainly. Tuition, program length, intake dates, admission requirements, co-op availability — on the page, in words and numbers, current. Brochure prose about transforming futures gives a language model nothing to work with. A clear answer to a real question gives it everything.

Keep one set of facts everywhere. Your website, Wikipedia, the directories, the provincial listings. When sources disagree, you've handed the model a coin to flip.

Structure pages so a machine can read them. One clear heading per question, a real answer under it, dates on anything that changes. This is the same discipline good SEO always asked for. The audience just got less forgiving.

Refresh what you can influence. That five-year-old news article ranking high in your sources list, the stale agent page, the abandoned FAQ — some of it you can update, some you can outrank with better pages of your own.

As a Canadian institution in 2026, you need to know what these tools say about you before your applicants do. The answers are sitting in anyone's phone, five seconds away. Go ask.