Someone Finally Read Your Whole Website.

The purpose of your web content hasn’t changed. Prospective students want to know the same things they’ve always wanted to know, and your institution is still the one place that knows the answers. 

What changed is that they’ve stopped asking you. They ask AI instead, from the couch, in full sentences. It answers in full sentences too. Few university contact forms have managed as much. That forces a question most of us have been able to put off for a long time. What is a university site for? I think I know the answer, at least for the next few years. 

Students still ask the same four questions. 

Nielsen Norman Group has been watching prospective students use university sites since 2014, and it turns out that whatever else changes about students, and a great deal does, they want to know the same four things. The latest version of the list came out in July 2022. ChatGPT came out that November. 

Do you have my program? Can I afford it? Will I get in? Will I belong? 

I’ve built information architecture around that list. I’ve mapped navigation to it, and I’ve used it to settle arguments about the top nav, or at least to shorten them. The questions are still right. Three of them now get answered by AI. 

Every conference keynote for the past three years has made that point, so I don’t need to make it again. What I hear much less about is what it reveals about how we’ve been storing the answers, and what it does to the fourth question on the list. 

What is a web page, anyway? 

A web page is simulated paper. Tim Berners-Lee built the web at CERN so physicists could share documents, and it arrived on computers that were already pretending to be offices, with desktops, folders, files, even a trash can. So we got pages.

A viewbook has a page 14, and the number is on it. When the number changes, you reprint the viewbook, and everyone understands the old one is now wrong. We carried that model online without really deciding to, and for three decades it worked well enough. The only concession to the medium is that web pages scroll, which technically makes them scrolls. 

Take six-year graduate earnings. At most institutions that number was typed by hand into a program page. Then typed again on an admissions page. Then again on a viewbook landing page, by an intern who has since graduated. Nobody in that chain did anything wrong. Each of them was handed a page and asked to put a number on it, which is what pages are for. 

So now there are three copies on three schedules. One gets refreshed when the new College Scorecard data comes out. The other two don’t, because nobody told their owners, and there was no reason anyone would have. This was always a bad idea. Institutions got away with it because a university site runs to tens of thousands of pages, many of them kept for reasons nobody remembers, and no one reads the whole thing. AI does. It finds all three numbers and picks one. 

A page is still a good place to read a number. It’s a bad place to keep one. 

This usually gets framed as getting AI-ready. That framing is too small. (It also sounds like a webinar title.) Tuition, outcomes, admission requirements, and program information are institutional knowledge. They should be kept the way the registrar keeps course credits, in one record with somebody’s name on it. 

The idea already has a name. In 2009, NPR called it COPE, for create once, publish everywhere. It took a radio network, whose main product has no pages at all, to notice. Most universities have settled on the other meaning of the word. 

Melvin Conway pointed out in 1968 that a system ends up shaped like the organization that built it, and a university site is the org chart with a search bar. Admissions holds its pages. Financial aid holds its pages. Each college holds a subdomain it seized during a redesign nobody else remembers. The earnings figure exists three times because three offices each wanted a copy they controlled. Keeping it in one record means two of them give up territory, which on most campuses is harder than moving a parking lot. 

The software part is solved, and has been for years. Almost nobody uses it this way, because it’s hard to get a budget for something no one can see. A redesign has a launch date and a reception with ice cream cake. Deciding which office owns the earnings figure has a meeting, and then another meeting. 

The best number of clicks is zero.

For fifteen years the discipline I work in has been organized around findability. In practice that meant shorter paths, cleaner labels, and fewer clicks to something. 

We got good at it. We ran card sorts and spent a week arguing about “Admission” versus “Admissions.” We shaved a click off the path to tuition and wrote it up as a case study. None of that was wasted, and none of it is enough anymore. 

Watch a student try to answer, “Can I afford this?” They land on the homepage. They look for admissions, or maybe financial aid, depending on who won the last reorg. They click down six or seven levels, past a PDF nobody has opened since it was uploaded. Every one of those steps is a tax on attention you spent real money to win. 

The old question was how to make that easier to find. The question now is why they should have to navigate at all. 

If someone knows their question, the experience should be able to answer it directly, on your domain, in your voice. Navigation has to stop being the only way to interrogate a university site. 

Accurate and forgettable, or beautiful and wrong.

A site now has to be two systems at once. One carries the facts. Call it the answer system. Its job is to be correct everywhere on the day a number changes, and nobody should ever notice it working. 

The other creates conviction. Call it the decision experience. It’s brand, story, pacing, and imagery — the sense that you’re somewhere specific and not inside a template somebody dropped a new logo into. Its job is to be remembered. 

Most universities build these one at a time, by accident. A redesign comes around every five years or so, and each one corrects for the last. The previous site was a CMS migration run by IT, so it was accurate and forgettable. The new one is a brand launch, so it’s beautiful, and the content migration was the first line item cut, so it’s wrong. Five years later the committee reconvenes and swings back. 

The way out is to stop making them compete for the same project. 

Nobody has ever felt at home in a chat window.

Information that anyone can retrieve in three seconds doesn’t set you apart. Your program list matters enormously, but it isn’t a competitive advantage. Your price might be. Publishing it isn’t, because AI will quote it next to four other schools in the same sentence. 

“Will I belong?” is a question about a feeling, and AI can’t retrieve it for you. 

AI can tell a student your student-faculty ratio is nine to one. But it can’t describe what nine to one feels like, which is a seminar room at four on a Tuesday in October, the radiator clanking, somebody’s coffee going cold, everyone having done the reading and the argument going well enough that nobody checks the clock. It can return your retention rate. It can’t produce the moment a seventeen-year-old looks at a photo of strangers on a lawn and decides, without quite knowing why, that they’d get along. 

That moment used to happen on campus. More students now put off the visit until after they’re admitted, when the tour is less a pitch than a victory lap. Until then your site is the campus, and they’re touring it from the couch, where they do the rest of their shopping. 

The feeling comes down to whether a story leads somewhere or just stops. Whether a faculty profile sounds like a person or a directory entry. Whether your sense of place shows up in the first fifteen seconds or sits three levels down under “Campus Map.” Whether a student could tell your homepage from the one forty miles away if you swapped the logos. 

Three of the questions are now better served as structured institutional knowledge, answered wherever they get asked, including places you don’t control and can’t see. The fourth one is yours alone, and it has become the most valuable thing your site can answer. 

AI can answer the facts. Only your site can answer what it feels like.