Somewhere in the wreckage of the “SEO is dead” discourse, a real story got buried. The Graphite/Similarweb study of 40,000 US websites found that the overall organic traffic decline of 2.5% was not evenly distributed. It was shaped like a squeeze, pressing hardest on the middle.

The top 10 sites by traffic grew 1.6%. The top 50 were roughly flat. Sites ranked 100 to 10,000, the vast middle tier of the web, took the heaviest losses. And below rank 10,000, results were mixed, with some small sites actually gaining.

This pattern tells you something specific about what Google is doing, and it is not “killing SEO.” It is concentrating organic traffic among fewer, larger, more authoritative domains. If you run a mid-tier content site, that distinction matters more than any headline about AI Overviews.

Who Sits in the Middle?

The sites ranked 100 to 10,000 by US traffic are a diverse group, but they share certain characteristics. Many are vertical content publishers: health information sites, recipe blogs, how-to guides, niche news outlets, comparison sites, affiliate content farms. They built their traffic by covering a broad set of informational keywords competently. Not brilliantly. Competently.

A site like this might rank for 50,000 keywords, hold positions 3 through 8 on most of them, and generate 2 to 10 million monthly visits. The content is accurate enough. The site loads fast enough. The backlink profile is decent. For years, that was enough.

Two things changed at once. Google’s AI Overviews absorbed some of the simple informational queries these sites depended on. And Google’s algorithm updates in 2024 and 2025 (the September 2024 helpful content update, the March 2025 core update, and the December 2025 core update) explicitly prioritized E-E-A-T signals: experience, expertise, authoritativeness, and trustworthiness.

The combination is lethal for mid-tier content sites. Their simple informational content gets eaten by AI Overviews. Their remaining rankings get outcompeted by larger sites with stronger authority signals. The squeeze comes from both directions simultaneously.

Why the Top 10 Grew

Reddit gained traffic. Wikipedia gained traffic. Amazon, YouTube, LinkedIn, major news outlets, government sites. The top 10 grew because Google’s algorithm increasingly treats brand authority and user-generated depth as ranking signals that override traditional content-based SEO.

Reddit is the clearest example. Google struck a data licensing deal with Reddit in early 2024, and Reddit pages now appear in roughly 60% of Google searches according to some tracking studies. Reddit’s content is messy, unstructured, and often wrong. But it is perceived as authentic. It contains real human opinions, which is precisely what Google thinks users want when they search for product recommendations, troubleshooting advice, or “best X for Y” queries.

A mid-tier comparison site writing “Best Running Shoes 2026” cannot compete with a Reddit thread where 400 actual runners argue about their favorites. The comparison site might be more accurate. It might be better organized. But Google has decided that the Reddit thread carries more E-E-A-T (specifically the “Experience” part) because it is written by people who actually bought and wore the shoes.

Whether this is a good decision by Google is debatable. The result is measurable: traffic is moving from mid-tier publishers to platforms where user-generated content provides perceived authenticity.

Why Small Sites Sometimes Win

Below rank 10,000, the results were mixed. Some small sites gained. This seems counterintuitive if the story is purely about authority (bigger = better). But the small sites that grew tend to share a trait: extreme specificity.

A site that covers one narrow topic with genuine expertise (say, a solo practitioner writing about specific woodworking techniques, or a researcher publishing detailed analysis of a niche ML subfield) can build topical authority that Google recognizes even without massive domain authority. These sites do not compete on breadth. They compete on depth in a space small enough that no large site bothers to cover it well.

Google’s algorithm has gotten better at distinguishing between “this site covers 10,000 topics adequately” and “this site covers 50 topics deeply.” The latter is favored. The former is the mid-tier content model that is losing ground.

The Content Model That Is Dying

Let’s be specific about what does not work anymore.

The programmatic content farm: a site that generates thousands of pages from templates, targeting long-tail keywords with thin, interchangeable content. These sites were already targeted by Google’s helpful content system in 2023-2024, and the subsequent updates have continued to demote them.

The “comprehensive guide” factory: a site that publishes 3,000-word articles on every conceivable topic, each structured identically (what is X, why X matters, how to do X, FAQ about X), with no original insight or first-hand experience. These articles were designed to be comprehensive for search engines, not useful for readers. AI Overviews now provide the same information more efficiently.

The affiliate content aggregator: a site that reviews products without testing them, compiling specifications from manufacturer pages and wrapping them in SEO-optimized prose. Google’s product review updates since 2021 have progressively demoted this content in favor of reviews with evidence of first-hand testing.

These models worked for years. They no longer work. The mid-tier traffic decline is concentrated among sites that relied on them.

What Mid-Tier Publishers Can Do

The squeeze is real, but it is not a death sentence. Mid-tier sites that adapt to the new environment can stabilize and even grow. The adaptations are specific.

Go deep on fewer topics. A site covering 500 topics at surface level will lose to Wikipedia, Reddit, and AI Overviews on every one of them. A site covering 50 topics with genuine depth, original data, and expert perspective can build the topical authority that Google rewards. Prune the shallow content. Double down on what you know better than anyone else.

Add experience evidence. Google’s E-E-A-T framework added “Experience” in late 2022, and the algorithm updates since then have progressively increased its weight. If you review products, show photos of yourself using them. If you write about techniques, show your own results. If you analyze data, show your methodology. First-person evidence is the one thing AI Overviews and Reddit threads cannot fabricate reliably.

Build off-site authority. Guest contributions to recognized publications. Conference talks. Podcast appearances. Citations by journalists. These are the signals that separate an authoritative niche site from a content farm covering the same keywords. Google’s algorithm uses off-site signals to validate on-site expertise claims.

Target queries AI Overviews cannot answer. “Best CRM for 3-person consulting firms in 2026” is a query where the user needs a detailed comparison, not a paragraph. “How to debug intermittent Kubernetes pod evictions in a mixed ARM/x86 cluster” requires technical depth that AI Overviews do not attempt. The queries that survived the zero-click wave are the ones with the highest commercial and professional value.

Accept the traffic loss on commodity content. Some traffic is gone permanently. The “what is X” article that ranked #4 for a high-volume informational keyword is not coming back to its 2023 traffic level. Redirect that content energy toward topics where you can be the primary source, not the fourth-best summary.

The Redistribution Is the Story

The 2.5% overall decline in organic traffic masks a redistribution that is reshaping the web. The top of the distribution is growing. The bottom is mixed. The middle is shrinking. That is not the death of SEO. It is the maturation of SEO into a channel that rewards genuine authority over mechanical optimization.

For mid-tier publishers, the honest message is uncomfortable: the strategy that built your traffic base is the strategy that is now losing it. The sites that adapt (fewer topics, more depth, real expertise, first-hand evidence) will find that organic search still sends traffic. More targeted traffic, from harder queries, to better content. The sites that keep publishing 3,000-word comprehensive guides on every keyword in the planner will keep losing ground to Reddit, Wikipedia, and the AI Overview box.

The middle of the web is not dying. It is being forced to earn its place.