Where Your Search Traffic Went

By Midya U · Midya U Advisory

In brief: Chartbeat data reported by Axios shows Google Search pageviews fell 34% for publishers between December 2024 and December 2025, hitting small, specialised sites hardest. AI tools are not absorbing that traffic — referrals from ChatGPT and similar sources still sit under 1% of pageviews. Your highest-intent program pages behave like small publishers and are losing the most. Rewrite them answer-first, before you commission anything else this quarter.

Someone on your team has pulled the analytics twice this quarter and got the same story both times. Rankings are holding, impressions are holding, and the clicks are down anyway. The explanation that gets written into the report is usually some version of "Google changed the algorithm again," and then everyone moves on to the campaign that has a deadline.

I would stop moving on. The numbers published in 2026 point at something more specific than an algorithm update, and the shape of it changes how you should write a program page.

Why is organic traffic falling when rankings are holding?

Chartbeat data shared with Axios in July 2026 put a figure on it. Pageviews from Google Search fell 34% between December 2024 and December 2025. Google Discover fell 16% over the same stretch.

The distribution is the part I would bring to a leadership meeting. Over two years, search referral traffic fell 60% for small publishers and 22% for large ones, with mid-sized sites at 47%. The penalty scales with how small and how specialised you are.

Your niche program pages behave like small publishers. A page about a two-year diagnostic imaging diploma has none of a national news site's domain authority behind it, and it exists to answer one specific question for a small number of people. That is precisely the kind of page an AI summary can absorb and repeat without sending anyone anywhere.

Is the traffic moving to ChatGPT instead?

This is the assumption I hear most, and the data does not support it. Chartbeat's numbers show AI referral sources sitting under 1% of publisher pageviews, even after ChatGPT referrals grew more than 200% year over year. Two hundred percent growth on a rounding error is still a rounding error.

The click is being absorbed at the point of the question and handed to nobody. Planning to make up your Google losses on AI traffic means planning around a channel that does not yet exist at any useful volume.

Which pages lose the most traffic to AI answers?

The ones that answer a question in a way a summary can lift cleanly. Tuition and fees. Admission requirements. Application deadlines. Whether the credential leads to a licence. Whether the program qualifies for a work permit.

Those are also the pages doing the most work in your funnel. A prospect who lands on your deadline page is much closer to applying than one who lands on your homepage. These losses hit at the bottom of the funnel, where each visit was worth the most.

One caveat about this read, because it matters. Chartbeat measured publishers, not universities, and to my knowledge nobody has published the equivalent curve for Canadian institutional domains. I am inferring across and it could be wrong in either direction. What I have not seen is an institution that can tell me which way it went for them.

What does this change about how you write a program page?

If a summary is going to lift your answer, the useful question is whether the answer it lifts is yours and whether it is right.

Write the answer before the persuasion. Put the tuition figure, the deadline, the prerequisite and the credential outcome in plain sentences near the top of the page, in the words a seventeen-year-old would type. Then make the case. Most program pages run this in the opposite order, opening with two paragraphs about "transformative learning experiences" before disclosing anything a person could act on, which reads fine to a human skimming and gives a language model nothing to quote.

What would I do this quarter?

Pick your twelve highest-intent pages. For each one, write down the question a prospect would type. Then put that question into Google, ChatGPT and Perplexity and record three things: whether an answer appears without a click, whether your institution is named in it, and whether the answer is accurate.

It takes an afternoon and one spreadsheet. Do it before you commission anything else this cycle, because it will change what you commission.

If you want to run this properly across the pages that carry your enrolment - what the AI assistants say about you now, where the answer is coming from, and what to rewrite first - that is what the AI Search Readiness Audit does.