AI Overviews squeeze organic traffic, reshape brand visibility

What happened

Per the HubSpot Marketing blog, the question "Is AI killing web traffic?" is no longer rhetorical. Google's AI Overviews, ChatGPT, Perplexity, and Gemini are absorbing the informational queries that used to send users to publisher pages. HubSpot's analysis pulls together what marketers have been seeing in their own dashboards: organic sessions falling on top-of-funnel content, even as keyword rankings hold steady.

HubSpot reports that the shift is structural, not seasonal. AI Overviews now appear on a growing share of commercial and informational searches, and when they do, click-through to the underlying sources drops sharply. Pew Research, cited in the same conversation across the industry, has put the click-through rate on AI-summarised results at roughly half that of traditional results.

The framing matters. HubSpot is not predicting the end of search traffic. It is describing a redistribution: fewer clicks per query, more brand exposure inside the answer itself, and a widening gap between sites that get cited in AI answers and sites that simply rank.

Why it matters for your brand

For B2B brands, the traffic story is the wrong story. The right story is citation share inside LLM answers. A CFO researching treasury platforms, a procurement lead at a multilateral comparing audit frameworks, a sustainability director scanning carbon accounting vendors: these people are increasingly getting their shortlist from an AI answer before they ever land on a vendor site. If your brand is not named in that answer, you are not on the shortlist. The click was never the point. The mention is.

This rewires content strategy for financial services in particular. Banks and asset managers have spent a decade building thought leadership hubs optimised for organic search on terms like "private credit outlook" or "Basel III implementation." Those pages still rank. They just get fewer human visitors, because the AI Overview at the top has already paraphrased the conclusion. The asset that now matters is whether Gemini, ChatGPT, and Perplexity quote your house view when a treasurer asks about rate exposure. That requires content structured to be extracted: clear claims, named authors, visible methodology, and distribution into the sources LLMs actually pull from.

For multilaterals and UN system agencies, the implication is sharper. Policy briefs and flagship reports were already competing for attention against think tanks and consultancies. In an AI-mediated information environment, the agencies that get cited are the ones whose data is machine-readable, whose findings are stated in extractable sentences, and whose authority signals (named experts, institutional URLs, structured data) are legible to a model. UNDRR's disaster risk data, for example, is more likely to be cited in a Perplexity answer about climate adaptation if it is framed as discrete claims tied to a credible institution, not buried in a 90-page PDF.

Industrial groups face a different version of the problem. A buyer researching low-carbon cement or grid-scale storage is asking AI tools for vendor comparisons. The AI answer often pulls from trade press, analyst notes, and Reddit threads, not from the manufacturer's own site. That means earned media and third-party validation now do double duty: they shape human perception and they feed the model's training and retrieval data. Holcim's visibility on "low-carbon construction materials" inside ChatGPT depends less on its own SEO and more on whether Reuters, ENR, and specialist publications are quoting it.