Semrush examined 200,000 AI Overviews — Google's AI-generated answer surfaces in U.S. search — and published the findings in a post titled "We Studied 200,000 AI Overviews: Here's What We Learned."

The 200,000-overview corpus, surfaced via the Semrush blog, ranks as the largest publicly disclosed dataset of its kind by a commercial SEO vendor (a company selling tools to search marketers). Semrush reported the headline figure itself, and the article indexed on Google News shortly after publication.

The dataset lands at a moment of intense publisher scrutiny. Coverage across the SEO trade press has tracked how AI Overviews affect referral traffic, citation patterns, and publisher economics — questions a 200,000-sample study could materially inform.

What are AI Overviews?

AI Overviews are Google's machine-written summary blocks that appear above traditional blue-link results (the standard list of webpage links) on Search. Google rolled the feature out broadly to U.S. users in May 2024, per company announcements, after a limited trial. The blocks use generative AI to synthesize answers from multiple web sources and link out to supporting pages.

Publishers and SEOs (search engine optimizers) have tracked the feature closely since launch. Trade-press reporting from outlets like Search Engine Land and Search Engine Roundtable has documented complaints about click-through declines on certain informational queries; Google has publicly stated that overall organic traffic remains healthy.

Why does a 200,000-overview sample matter?

Earlier independent analyses typically ran in the hundreds or low thousands of queries. Semrush's 200,000-overview corpus sits orders of magnitude above that prior benchmark — and would, if methodology holds, give the industry the largest published snapshot yet of how the feature behaves across queries.

That scale matters because AI Overviews vary widely by intent type. Consider how a brand-search query differs from a how-to query. Smaller studies struggle to surface those patterns across intent; larger ones can. The number Semrush cites suggests an attempt at a higher-resolution view.

What did Semrush actually publish?

The post's headline doubles as the vendor's framing: "We Studied 200,000 AI Overviews: Here's What We Learned." Semrush's research byline carries the piece, and it lives on semrush.com — a domain that has hosted SEO (search engine optimization) research since 2008. The company is best known for keyword research and backlink (inbound link) analysis, and it has expanded into content marketing and AI-search tracking.

What's confirmed vs. tool-signal speculation?

Confirmed so far: the study's existence, its title, its publisher, and the 200,000-overview sample-size figure — all directly attributable to Semrush. Tool-signal speculation, not yet verified in indexed excerpts: which queries were sampled, how the audits were collected, what vertical breakdowns will follow, and which Semrush researchers are quoted.

Until readers pull the post in full, third-party recaps should be treated as preliminary. The methodology section is expected to land on the same URL.

What's next?

Semrush has historically followed broad studies with vertical-level breakdowns and product integrations. A follow-up post splitting out YMYL (Your Money or Your Life — finance, health, legal) queries, e-commerce queries, or local-intent queries would track with that pattern. Conference talks at upcoming search-industry events are also plausible.

For now, the headline fact holds: someone ran a number on AI Overviews, and the number is 200,000.