Our blog team came to me with a problem and an assumption.

The problem: traffic was down, views were down, and the trend lines only pointed one way. The assumption: somewhere there was a technical fix, a setting we could flip, that would turn it back around.

They were partly right. That’s what made the conversation hard. Because the part where they were wrong is the actual story, and it’s one every marketing team in healthcare is about to live through, if they aren’t already.

TL;DR

The meeting

When we dug into the numbers, part of what the team was seeing was on us. We did plenty well: strong writing, real clinical expertise behind it, a steady publishing rhythm. But we had gaps. Our structured data wasn’t fully implemented. Our pages weren’t always formatted the way search engines reward, with the clear subheadings, the tight summaries, the depth on a topic that signals you actually know it. Looking back, some of our old traffic was never really earned. It came from a search engine that rewarded being present more than being good. AI raised the bar, and we weren’t clearing it everywhere we needed to.

So yes, there was work we owned. But fixing it wasn’t going to restore the old numbers, and I needed the team to understand why before we spent a year chasing a ghost.

We weren’t going to code our way out of this. We could mitigate the decline, not stop it — and neither could anyone else.

What we actually had to do came down to three things at once: build what was missing, teach the writers a new way to work, and lead the team through a change none of us chose. Not a fix. A reorientation.

It’s not just you

Here’s the part that isn’t anyone’s fault. The behavior of search itself changed.

When Google shows one of its AI Overviews (the AI-written answer that now sits at the top of the results), people stop clicking. Pew Research studied real browsing data and found that users click a traditional search result about 8% of the time when an AI summary is present, versus 15% when it isn’t. The AI answer roughly halves the click. And it doesn’t hand that traffic to the sources it drew from: links inside the summary get clicked about one percent of the time. More and more, the answer is the destination. Nobody leaves.

The penalty is growing, too. Ahrefs measured the click loss to the top organic result at 34.5% in early 2025; by year’s end, it had climbed to 58%. In under a year, the cost of being “number one” nearly doubled.

And healthcare is at the front of the line. AI Overviews now appear on 48.7% of health-related Google searches, the highest rate of any industry studied. Almost half. Meanwhile, the AI tools sending traffic back account for less than one percent of healthcare site visits. The clicks aren’t migrating to a new channel. They’re evaporating. None of this is a local failure of your search engine optimization (SEO). It’s the weather.

Sitting in that data, I finally named the thing we’d been dancing around: we were a publisher. A health system with a blog and a content calendar is in the same business as a newspaper, and the newspaper business was on fire. We’d just been slow to smell the smoke.

The anonymous impression

For a while I told myself the answer was obvious. If the click is going away, win the citation. Be the source the AI quotes. There’s a whole cottage industry forming around exactly that: “generative engine optimization,” “answer engine optimization,” a fresh acronym every quarter. Google itself has mostly waved it off. The company’s own guidance says optimizing for its AI features is really just SEO by another name, and that the special AI-only markup people are selling does nothing.

But here’s what stopped me. Even if you win, even if the AI names your article as a source, you mostly win anonymity. Remember that one percent. Your content shaped the answer, the patient got helped, and your brand got an impression no one saw and no one can trace. In the old world, a top result meant a click, a visit, your name in front of a person, a shot at whatever came next. Now it’s a ghostwritten sentence in someone else’s voice.

That’s not a win. It’s an anonymous impression. And it raises the question this whole piece is really about: when the click disappears, how do you still get credit for helping the patient?

When the most measurable channel becomes the least

Step back and the shift is bigger than traffic. It’s measurement.

For twenty years, digital marketing sold itself on one promise: everything is trackable. Every click, every source, every step on the path to a conversion. We built teams, budgets, and whole belief systems on that promise. We were trained to chase what we can measure and to distrust what we can’t.

AI search breaks the promise. When the answer arrives inside a summary and the source goes unclicked and untraceable, the most accountable channel we ever had starts to behave like the least. You influenced a decision. You can’t prove you did.

Digital was the most measurable channel we ever had. AI search is quietly turning it back into radio.

Radio worked. Billboards work. Plenty of great brands were built on channels nobody could tie to a sale. But it takes a different temperament, a tolerance for influence you feel more than you count. For a generation of marketers raised on dashboards, that’s the real adjustment. It’s harder than any technical one.

What actually works

So what do you do on Monday? Three things, and not one of them is a plugin.

Tell the stories only you can tell. The instinct is to double down on clinical content: condition pages, symptom explainers, treatment overviews. Resist it. Be honest with yourself, because clinically, most health systems look alike, and an AI can assemble a competent symptom answer from a dozen sources without ever needing yours. What it can’t source anywhere else is your stories. The patient who got her life back. The nurse who’s worked the same unit for thirty years. What your team did when the community needed it. Testimonials, people, local impact: these are unique to you by definition, and no model can synthesize them from someone else’s site. That’s where I’d put the weight.

Own the next step. The clinical question gets asked and answered first, maybe by an AI, maybe anonymously. Fine. Watch what comes right after it: “where do I get treated for this?” and “who’s the best doctor for that?” Those are the conversion questions, and they’re yours to win. You might lose the impression on the diagnosis and still become the default answer to where and who, which is the part that was ever going to fill a schedule anyway.

Learn to experiment at scale. For years the gold standard of healthcare content was an army: strategists, researchers, writers, editors, videographers, designers, all feeding one machine. Cleveland Clinic built a famous one, and it was a real moat. That moat is draining. AI means almost anyone can create, test, and iterate content across formats and messages at a volume that used to take a department. That isn’t a reason to gut the team; it’s a reason to turn it into a lab. The teams that win the next few years are the ones that experiment fastest, because nobody actually knows what works yet, and the only way to find out is to go first.

Don’t wait for the cavalry

Fair question: won’t someone stop this? The lawsuits, the regulators?

I wouldn’t build a plan on it. The publisher lawsuits against the AI companies are mostly resolving into licensing deals, not injunctions. Courts are leaning toward “training on content is fair use, so pay for it,” which keeps the AI answers coming. Regulation is the real wildcard. The UK just became the first to force Google to let publishers opt out of its AI Overviews without losing their search ranking, and the EU has opened an investigation of its own. Rules like that might slow the bleed or buy some time. They won’t put the click back the way it was. Plan for the world that’s arriving, not the one you’re hoping gets restored.

My bet, on the record

I’ll say it plainly, because a prediction you can’t be wrong about isn’t worth making: I think organic search as we’ve known it is largely finished by late 2028, with non-AI results reduced to a low-single-digit share of how people actually find answers. The curve is still accelerating. It’ll slow before it’s done, but the direction isn’t in doubt. Analysts are already pointing the same way; Gartner expects traditional search volume to fall 25% by 2026 as AI assistants absorb the queries.

I could be wrong, and I’ve been specific about what would change my mind: meaningful legislation or litigation, AI baked into our phones at the operating-system level, a real rival to Google taking serious share, or a public backlash against AI answers that sends people back to links. I put all of it on a clock, and I’ll move the clock in public as the signals move.

What we were really after

The blog team wanted the old numbers back. I get the impulse; I felt it too. But the click was never the thing we were actually after. It was a proxy, a countable stand-in for the real goal, which was to help someone and be remembered for it.

Search spent two decades teaching us to be findable. The next decade is about being worth seeking out. The click is ending. The relationship doesn’t have to.