'SEO Is Dead' Is Lazy. 'Nothing Has Changed' Is Just as Lazy.

Both camps are selling something.

The people announcing the death of SEO are usually selling the replacement. The people insisting nothing has changed are usually selling the retainer that would be harder to justify if something had. Neither position requires evidence, which is why both are so popular.

The evidence that does exist is narrower and more interesting than either camp wants it to be.

What the data actually says

In March 2025, Pew Research Center tracked the real browsing behavior of 900 US adults across 68,879 Google searches. Of those, 12,593 returned an AI summary.

The finding: users who encountered an AI summary clicked a traditional search result in 8% of visits. Users who did not encounter one clicked a result in 15% of visits. Nearly double.

A second number is worth sitting with. Clicks on links inside the AI summary itself happened in about 1% of visits. The citation was there. Almost nobody used it as a door.

You can read the whole study at Pew Research Center.

What it does not say

This is where I part company with most of the people quoting it.

It is observational, not causal. AI summaries appear more often on certain kinds of queries, and those queries may have had lower click rates anyway. The study cannot separate the effect of the summary from the effect of the question.

It is US adults, one month, one search engine, in early 2025. It is not a global figure, it is not a 2026 figure, and it is certainly not a forecast.

It does not segment by commercial intent, which is the segmentation that would actually tell you what happens to your revenue. It does not measure conversions at all.

So anyone converting 8% versus 15% into a prediction about your pipeline is overreaching. And anyone using those caveats to conclude that nothing is happening is doing something worse, because the direction of the finding is not really in dispute and it lines up with what most operators are seeing in their own dashboards.

The honest reading

Discovery can stay valuable while producing fewer visits.

That single sentence resolves most of the argument. Your content can do its job, reach the person, shape the decision, and never generate a session. The value did not disappear. It moved to a place where your analytics cannot see it.

This is uncomfortable for an industry that has reported sessions as its headline number for twenty years. It is not an existential threat to the work. It is an existential threat to the reporting.

What to change in how you measure

Stop leading with sessions. Organic sessions are now a partial measure of a partial funnel. Keep the number, demote it.

Split branded and non-branded, always. If AI answers are doing their job, some people who would have landed on a comparison page now arrive by searching your name directly. That shows up as branded search growth and direct traffic, not as organic growth on the page that did the work. If you report those separately you can see the trade. If you do not, it looks like decline.

Watch impressions against clicks. A widening gap in Search Console between stable impressions and falling clicks is the signature of this shift. It is a much earlier signal than a traffic drop, and it tells you which page templates are exposed.

Segment pages by what they do, not by folder. Informational content that answers a question completely is the most summarizable thing you own, and should be expected to lose clicks. Pricing, tooling, comparison, and anything requiring a logged-in state is far more resilient. Report those cohorts separately or the average will hide both stories.

Ask people where they came from. A single open field on your demo form, asking how someone heard about you, will tell you more about assistant-driven discovery in a quarter than any dashboard currently on the market. It also keeps you honest about the difference between being cited and being bought.

What to change in what you build

The pages that survive this are the ones a paragraph cannot replace.

If the entire value of a page is a fact, a summary will deliver that fact and the visit was never necessary. If the value is a calculation with your inputs, a comparison built on data you collected, a tool that does the task, or access to something only your product has, there is nothing to summarize. The answer has to send the person to you.

That is not a new insight dressed in AI language. It is the same instinct that made good pages good before any of this. The difference is that the penalty for ignoring it used to be mediocre rankings, and now it is being quietly answered on your behalf.

The position worth holding

Something real changed, the size of it is not yet settled, and the responsible move is to change your measurement before you change your strategy.

Anyone certain in either direction is guessing. The 8% figure is a strong reason to take the change seriously. It is not a reason to panic, and the caveats around it are not a reason to do nothing.