AI Search Traffic Bleeds Publishers but Feeds Homepages

On May 7, ChatGPT changed how it showed clickable brand links, and AI search traffic immediately revealed the web’s new contradiction: machines cite deep pages, but humans increasingly land on the homepage.

That timing matters because the old bargain behind search is breaking. Publishers and brands fund pages that AI systems read, quote, and summarize. But the human clicks that used to pay for that work are shrinking. The result, as VentureBeat argues, is not simply “AI is killing the web.” It is stranger. AI needs the open web to stay useful, while its answer boxes make the open web harder to fund.

XOOMAR analysis: the old click economy is not recovering to its prior shape. The winners will be sites that split their architecture in two: deep pages built to be cited by machines, and front-door pages built to convert humans who arrive already informed.

May 2026 data shows AI search traffic moving against publisher clicks

Pew Research Center tracked browsing behavior from 900 U.S. adults and found a sharp drop in traditional clicking when Google shows an AI summary. Users clicked a standard search result 8% of the time when an AI summary appeared, compared with 15% when it did not. Links inside the AI answers performed worse, drawing clicks only about 1% of the time.

Publishers are feeling that math. Chartbeat data reported by Axios showed page views from Google Search fell 34% across its publisher network between December 2024 and December 2025. Small publishers have lost roughly 60% of search referral traffic over two years. Business Insider’s organic search traffic dropped 55% over three years, and some smaller publishers have already shut down.

At the same time, machines are reading more. Similarweb’s 2026 generative AI report found that the share of ChatGPT answers with live web citations grew more than fivefold in under a year, reaching 6.8% of all answers by May 2026. In travel, that figure hit 22.6%.

That is the core fracture. Human visits fall. Machine consumption rises.

If organic visibility dips, AI search visibility follows, because models are less likely to find the content.

That point, attributed in the source material to Lily Ray, VP of SEO and AI search at Amsive, cuts against the idea that publishers can abandon traditional search work and simply “optimize for AI.” The retrieval layer still depends on findable, structured, current pages.

But chatbot referrals remain tiny. They still account for less than 1% of publisher page views, even after growing more than 200% in a year. AI search traffic is growing, but it is not replacing the volume Google Search removed.

May 7 turned ChatGPT referrals into homepage traffic

After ChatGPT’s May 7 search update, which surfaced more prominent clickable brand links inside answers, referral traffic from ChatGPT surged 157% in a week. The more revealing shift was where those visitors landed. Homepage landings more than doubled, from roughly 25% to nearly 60%.

That is a different referral model.

Traditional search sent users into the page that matched the query. AI systems increasingly do the comparison first, then send users to the brand’s front door. The cited page and the clicked page are no longer the same asset.

The split is visible in the data. Similarweb found 65% of ChatGPT-cited URLs sit two or three folders deep in a site, while 58.8% of referral traffic lands on homepages. Ahrefs saw more than 80% of its AI referral traffic go to its homepage, product pages, and free tools, not its large editorial library.

A third destination is emerging too. Previsible analyzed 6.77 million AI-referred sessions and found 28.8% of ChatGPT referrals land on internal site search pages. That turns a neglected UX feature into an acquisition surface.

XOOMAR analysis: most sites are still built for the old flow. They assume the article is both the evidence and the entrance. AI search traffic breaks that assumption.

December 2025 put Google’s tollbooth under legal and commercial pressure

Google is not watching this from the sidelines. It is both the incumbent losing clicks and a major force pushing AI answers into search.

AI Overviews appeared in more than 40% of Google searches by May 2026, per Similarweb. Visits to Google’s conversational AI Mode have climbed since launch. That means Google is cannibalizing some of its own click economy because the alternative is letting OpenAI, Perplexity, and Microsoft train users to expect answers instead of result pages.

The pressure is also commercial. eMarketer projects Google’s share of U.S. search advertising will fall below 50% in 2026, the first time since roughly 2004. The biggest chunk of lost share is going to Amazon, whose sponsored product searches count as search advertising and are growing three times as fast as Google’s.

There is a legal layer as well. A federal court entered final judgment in the DOJ search antitrust case in December 2025, imposing remedies that bar exclusive default agreements and require Google to share search data with qualified competitors. Google appealed in January 2026. The DOJ cross-appealed seeking stronger remedies.

XOOMAR analysis: Google’s strategic bind is clear. It has to protect search advertising while adapting to a format that reduces the need to click search results. That sits beside broader AI platform pressure covered in our reporting on Microsoft AI Models Drag OpenAI Into a Margin Fight, and Amazon’s separate push into AI-driven business workflows in $200 Billion Sales Let Amazon AI Agents Invade Workflows.

June 2026 showed ads moving inside the conversation

Money is following the user behavior. Sponsored results appeared in 26% of U.S. desktop ChatGPT conversations in June 2026, up from 14% one month earlier, according to Similarweb ad intelligence data cited in the source material. Two-thirds of those ads appeared after the second prompt, targeted on conversation context rather than a keyword. Click-through sat around 0.50%.

That is not the old keyword auction with a chatbot skin. It is paid placement inside a live exchange, where the ad can appear after the system has collected more context about the user’s intent.

For publishers, that creates an uncomfortable outcome. Their reporting, reviews, comparisons, and explainers can feed the answer. The ad value may then accrue to the platform or to the brand placement inside the chat, not to the page that supported the response.

For brands, the incentive is different. AI referrals may be smaller in volume, but Similarweb data shows AI-recommended brands receive two to four times as many subsequent visits as competitors that were not recommended. A smaller stream of higher-intent visitors can matter more than raw sessions.

For SEO teams, the playbook gets messier. Ahrefs found 67% of ChatGPT’s most-cited sources are things marketers cannot influence, with Wikipedia alone accounting for nearly 30%. It also found 28.3% of ChatGPT’s most-cited pages have zero Google organic visibility, which weakens any simple claim that classic SEO rankings translate cleanly into AI citation.

Sites now need one layer for citation and another for conversion

The practical response is not to chase every AI referral dashboard. It is to separate page roles.

Deep pages should be built for citation. That means specific claims, clear headings, descriptive URLs, comparison tables, benchmarks, and visible provenance. Ahrefs found pages with natural-language URL slugs get cited at 89.78% versus 81.11% without.

Homepages need to serve a visitor who has already been briefed by a model. They should answer: what do you do, who is it for, why trust you, and what should the visitor do next. Fast.

Internal search deserves real investment because AI systems are sending users there. If nearly a third of ChatGPT referrals in Previsible’s dataset land on internal search pages, that page is no longer just a utility. It is part of acquisition.

The next evidence to watch is whether AI referrals remain small but high-intent, whether homepage and tool-page conversion keeps rising, and whether publishers can create a funding model that pays for the pages machines rely on. If those signals strengthen, the thesis hardens: the open web’s next winners will be both quotable by machines and easy for ready-to-act humans to use.

The Bottom Line

• AI summaries are reducing the search clicks publishers rely on to fund content.
• Brands may need separate strategies for machine-readable deep pages and human-focused homepages.
• The open web faces pressure because AI depends on publisher content while sending fewer readers back to it.