Is SEO actually dying? Here's how to tell for your site

You know the whiplash. Monday morning, someone posts a chart showing organic traffic is fine, actually — down a couple points, nothing to see. By Wednesday, a writer whose site cratered is on your feed asking whether their career survives the year. Both posts have data behind them. Both are describing the same twelve months, the same search engine, sometimes the same industry. The question everyone keeps asking — is SEO dying — turns out to be the wrong shape for an answer. It isn't a yes-or-no question. It's a where-do-you-fall question, and this week gave us enough real numbers from both camps to actually answer it for your own site instead of arguing about it in the abstract, again, on a group chat that never resolves anything.

Is SEO dying? The optimists' data says no

Start with the camp that says calm down. Ethan Smith, CEO of Graphite, published a rebuttal to the doom narrative built on Similarweb data across more than 40,000 U.S. sites. His finding: Google organic traffic is down about 2.5% year-over-year. Not 25%. Not the 60% figures that have been getting passed around in some corners of the industry. Two and a half points, in aggregate, across a sample large enough that outliers shouldn't be steering the average.

If you're trying to make sense of the organic traffic decline headlines populating your feed, the size of Smith's sample matters. His piece was a direct response to specific claims — 25%, 40%, 60% drops — that had been circulating without much sourcing behind them. When he pulled a dataset spanning tens of thousands of domains, the number that came back was small enough to call flat. If you're building a SEO program on the assumption that organic search is a dying channel across the board, that assumption doesn't survive contact with this dataset.

This matters for anyone deciding a budget, not just anyone winning an argument. A leadership team that reads '60% decline' and panics might pull spend out of a channel that, in aggregate, is still doing roughly what it did last year. A team that reads '2.5% decline' and relaxes might miss that the number is only calm because their own site is being offset by someone else's gains — which is a very different problem, and one the aggregate figure can't tell you about on its own.

It's also true that the numbers underneath that flat aggregate have shifted more than the average lets on. Research from Smith and Graphite, covered by Search Engine Land in March, found that AI assistants now account for roughly 56% of global search engine query volume — something like 45 billion monthly AI sessions. ChatGPT alone is responsible for about 89% of that AI session volume. And Google's own share of total search volume fell from roughly 89% in 2023 to about 71% by the fourth quarter of 2025. So query volume is shifting hard toward AI assistants even while the traffic Google sends to the open web stays close to flat. Those two facts coexist because AI search traffic and organic traffic aren't measuring the same thing — one measures where people search, the other measures where they land.

Is SEO dying? The Google Zero stories say yes

Now the other camp, and their evidence isn't abstract. Forbes ran a piece in May on what people have started calling 'Google Zero' — the point at which AI-generated answers absorb enough of a page's clicks that direct search traffic effectively disappears. The piece centers on Nicholas Bouliane, who runs All About Berlin, a site built over years on detailed relocation guides. After AI search rollouts, his traffic dropped 70%.

“Google broke the economics of putting out free information. The damage to the independent web is incalculable.”

That's Bouliane, describing what happened to a site he spent years building. He isn't alone in that read. Former Washington Post editor Karen Attiah told Forbes, "Google is about to kill everyone's website, basically" — a sentence blunt enough that it doesn't need much unpacking.

The same piece notes that People Inc., one of the larger publishers navigating this shift, has watched its search-referral traffic share fall from roughly 65% three years ago into the high-20s percent range now. That's not a rounding error. That's a business model getting rebuilt in real time, and if your revenue or your paycheck depends on search-referred pageviews the way a magazine's does, a drop like that isn't theoretical.

It's worth sitting with why these stories land the way they do. Bouliane and Attiah aren't cranks and they aren't exaggerating for clicks — they're describing something that actually happened to sites they know intimately, built over years of real work. Dismissing that because a separate dataset looks calmer would be its own kind of dishonesty. The stories are true. They're just not the whole picture, in the same way Smith's average isn't the whole picture either.

So both camps have real numbers. Smith isn't wrong about the 2.5%. Bouliane isn't wrong about his 70%. The mistake is assuming only one of them can be telling the truth.

Why both sides are right

Here's the idea that resolves it, and it's simpler than the argument makes it sound: an average can stay flat while the sites inside it move in opposite directions. If a meaningful share of the web is gaining traffic and a meaningful share is losing it in roughly equal measure, the aggregate number looks calm. That calm isn't evidence that everyone is fine. It's evidence that the losses and the gains are canceling out in the math while doing nothing of the sort in anyone's actual business.

Cyrus Shepard, founder of the SEO firm Zyppy, published research in April that gets underneath that average. He analyzed more than 400 sites that had won or lost traffic and looked for what separated them. Five traits came out of that analysis, and the gap between winners and losers on each one is not subtle.

Stack enough of those traits together and the pattern turns stark. Sites with four or five of the five traits won — meaning they gained traffic — 68 to 70% of the time. Sites with none of them won only 13.5% of the time. That's the mechanism behind the flat average and the horror stories, at once. A site built almost entirely on free informational guides, with no product to sell and no proprietary data behind it, sits close to the losing profile Shepard describes. A site with a real product, a tight topic, and its own data sits in the profile he found protected — and its traffic gains show up quietly in Smith's average, pulling the number back toward flat while someone else's site free-falls in the same period.

It's also why generative engine optimization keeps coming up in the same conversations as classic search work rather than replacing it. The traits Shepard found protect a site in Google's rankings — proprietary data, task completion, topical depth — are the same assets AI answer engines pull from when they decide who to cite. Building for one increasingly means building for both.

That also explains why the 'pick a side' framing of this debate never quite works. Teams that treat it as an either/or question end up either overinvesting in a channel that's genuinely shrinking for them, or underinvesting in one that's still carrying real weight. The more useful move is to stop asking which camp is correct and start asking which camp your own analytics agree with — which is a data question, not an opinion question, and one you can usually answer in an afternoon with a spreadsheet and twelve months of Search Console history.

So is SEO dying, or is it just your site?

This is where the debate stops being interesting and starts being useful, because Shepard's five traits turn into a five-minute audit instead of a philosophical argument. Pull up your own site and score it honestly against each one.

Score yourself. Zero or one of five, and the 'SEO is dying' headlines are describing you specifically, not the industry — worth treating as urgent. Four or five, and you're closer to Smith's dataset than Bouliane's, even if the discourse doesn't bother distinguishing between the two of you.

If tight topical focus is your gap, topical authority signals in AI search is a reasonable place to start — it's the same idea Shepard's data rewards, worked through in more detail. If you can't tell whether your own traffic swings are a Bouliane problem or a Smith problem, attributing pipeline to organic and AI search walks through separating the two in your reporting instead of guessing from a headline. And if you want a concrete example of a site that built the kind of proprietary-data moat Shepard's research rewards, the Marketcheck case study is worth a look.

None of that requires abandoning search as a channel, and none of it requires believing either camp is lying. It requires the unglamorous work our content marketing team spends most of its time on: building something with an actual product behind it, a job it finishes on the page, and a subject it owns instead of skims. It's also slower than either headline suggests — nobody flips from zero traits to five in a sprint, and treating this like a fire drill tends to produce the same thin, everywhere-at-once content that put a site on the losing side of Shepard's data in the first place.

So — is SEO dying? Wrong question, and it always was. The right one is whether your site looks like Ethan Smith's average or like Nicholas Bouliane's outlier, and that's not a mystery you have to wait on the next hot take to resolve. Run the five-trait check this week. You already have the answer sitting in your own analytics.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.