The Trust Gap AI Hasn't Closed

By Midya U · Midya U Advisory

In brief: Keystone Education Group's 2026 survey of more than 67,000 prospective students found university websites trusted more than twice as much as AI. Students are routing through AI to find you and then coming to your site to check whether it's true, which means the visits that survive matter more, not less. This piece covers what those students check first, and the ten-minute audit to run on your highest-traffic pages.

The sentence I keep hearing in 2026 runs roughly like this: students don't come to the website anymore, what are you doing about AI. It usually arrives from someone senior who read something about AI search earlier in 2026, and it's a fair conclusion to draw from a traffic report.

The survey evidence points the other way, and it has changed what I tell people about their websites.

Do students trust what AI tells them about universities?

About a third of them do.

Keystone Education Group released its State of Student Recruitment 2026 in May 2026, drawn from more than 67,000 prospective students in over 150 countries, surveyed between 1 October 2025 and 30 April 2026. Students were asked which sources they use weekly or daily, and separately which ones they trust very or extremely.

• University websites: used by 59%, trusted by 72%
• Web search: used by 73%, trusted by 51%
• Social media: used by 58%, trusted by 27%
• AI: used by 51%, trusted by 30%

So half of them are asking AI, and under a third believe the answer. Keystone's own copy describes that as "nearly 4x more trust" in websites; the chart it sits above reads 72 against 30, which is closer to two and a half times. The gap is still the finding. I'd quote the two numbers and leave the multiplier alone.

If AI is cutting our traffic, why does our website matter more?

Because the visits that survive are the ones sitting closest to a decision.

I wrote in August 2026 that publisher pageviews from Google Search fell 34% between December 2024 and December 2025, and that the missing clicks were being absorbed at the question rather than handed to another channel. That still holds, and it sits alongside this one without contradiction. Fewer people arrive, and the ones who do have already been told something about you by a source they half-believe. They have come to check it.

Where that first answer comes from, and how much of it you can influence, is a different problem with its own fix, and I've covered it in what an assistant says to a 17-year-old. This piece is about the ten minutes after they click.

What do students actually check before they apply?

Money and outcomes, mostly.

The same report ranks decision factors by weighted score. Cost sits at 87, graduate outcomes at 84, internships at 83, accommodation at 82, and the profile and quality of academic staff at 81. Rankings come in at 73. History and heritage finishes last, at 69.

Most institutional websites lead with heritage and a ranking, and reach the fee schedule somewhere past the second scroll. That is the report's list upside down.

Email is how 63% said they'd prefer to talk to an institution, against 4% for live chat on the website and 3% for social media. UniQuest, a Keystone company, adds that 85% of students expect a same-day reply. Your program page needs to hand them a real email address and a person behind it who answers that day.

Does this hold for domestic Canadian applicants?

Keystone surveys its own audience: prospective international students who found their way onto study-search platforms and are already comfortable researching a degree online. A grade 12 student in Surrey choosing between two BC institutions, with a parent in the room and a campus she can drive to on a Saturday, is a different person carrying her own set of worries. I'd expect the direction to hold, and I wouldn't transfer the size of that trust gap to a domestic funnel without testing it first. If your inquiry data says something about this, you're holding evidence the sector doesn't have.

What should we fix on our website first?

Take your ten highest-traffic program pages and read them as someone who has just been handed your name by ChatGPT.

On each one, time how long it takes to find four things: the total cost for one year, the entry requirements, what graduates of that program are doing now, and the name of a human who will answer an email. Anything over about fifteen seconds is a page written for an audience that already trusted you.

Then check the date each page was last edited. In the audits I run, that column usually tells me more than the traffic chart does — the pages carrying the most decision weight are often the ones nobody has opened since the market changed underneath them.

Rewriting those pages around how applicants decide is what my Brand Repositioning engagement does. The parent-survey evidence in The Case You Stopped Making for a Degree points the same way.