Semrush has published a study titled "AI Overviews Study: What 2025 SEO Data Tells Us About Google's Search Shift." The announcement surfaces through Google News aggregation. The visible release, at the time of writing, contains no figures, no sample sizes and no methodology.

That last sentence is the story right now. A study without published numbers is a claim, not a finding.

What has Semrush actually confirmed?

Only the existence of the study. The title tells us three things:

• The data covers 2025.
• The subject is Google's AI Overviews.
• Semrush frames the findings as evidence of a "search shift."

Nothing else is verifiable from the current release. Semrush has not stated the number of queries analyzed, the date range of the crawl, or the geographic markets covered.

Why this matters anyway

Semrush sits on one of the largest clickstream and rank-tracking datasets in the industry. When the company publishes on AI Overviews, practitioners read it closely. Prior tracking from Semrush and competitors such as BrightEdge and seoClarity has shown AI Overviews expanding into more query categories through 2024 and 2025.

But scale of data does not guarantee rigor of analysis. The caveat stands until Semrush publishes its sample sizes and method.

What is not said

The release does not say which verticals the study covers. It does not say whether AI Overviews raised or lowered organic click-through rates. It does not distinguish desktop from mobile behavior. And it does not state whether Google confirmed any of the trends the tool data suggests.

That distinction matters. Tool vendors measure what their panels show. Google confirms what its systems do. These are different evidentiary classes, and SEO coverage too often blends them.

Takeaway: read the study when the full dataset lands, and grade every claim by whether it carries a sample size.

What to monitor next

Watch the Semrush blog for the full methodology section. Compare any click-loss figures against Google's own statements on AI Overviews traffic, which have historically disputed third-party loss estimates. If the study names affected verticals, cross-check them against your own Search Console query data before drawing conclusions.