Human web traffic is declining in measurable, accelerating ways. Cloudflare data captured in a report by CBC.ca found that retail, software, IT, and finance websites saw human visits fall by as much as 40 percent between June 2025 and April 2026. Over the same period, Google rolled out AI Mode, a conversational reimagining of its search box, and that product crossed one billion monthly users. The machine-reader shift is not approaching. It has arrived.

AI Intermediaries Are Absorbing Queries Before Users Click

The mechanics are straightforward. Google is now prioritizing extensive AI-generated summaries over links to publisher pages, which means users increasingly retrieve answers without leaving Google’s main page at all. AI Mode represents the first fundamental redesign of the Google search box in 25 years, and its scale suggests this is not a transitional experiment.

For publishers, the consequences are concrete. Food blogger Deb Perelman, whose site Smitten Kitchen marks its 20th anniversary this summer, has watched her recipe content appear inside AI summaries she never authorized.

“It’s honestly shocking to me that somebody can publish my work … without my permission.”

Perelman is now pursuing AI-limiting protections for her site. Her situation reflects a structural problem that extends far beyond individual recipes. When an AI summarizer extracts and presents content, it removes the context a human reader would encounter on the original page.

“The context of a recipe is really important. You’re not going to get that from a blurb on top of a search result that has been extracted, perhaps correctly, perhaps incorrectly, from the actual recipe.”

Her case illustrates what publishers across categories are absorbing: an intermediary has inserted itself between their content and their audience.

Real-Time Data Publishers Face the Sharpest Structural Pressure

The editorial team at Live Sports Odds, which tracks how fast-updating data pages are built and surfaced, observes that the Cloudflare traffic declines and the rise of AI intermediaries land with particular force on publishers whose pages serve high-intent, real-time data queries. For a sports news site, a delayed summary may inconvenience a reader. For a publisher whose audience needs precise figures in the moment, an inaccurate or outdated AI summary causes direct harm.

Cricket betting odds represent one concrete case of this problem. A provider of live cricket betting odds must re-engineer its page architecture and metadata so an AI summarizer can retrieve fast-changing numbers with accuracy, not present figures from a prior session as current. The challenge is structural: when AI intermediaries intercept queries before a user reaches the page, the publisher cannot control what the summarizer surfaces unless the underlying data is formatted for precise machine retrieval.

“The biggest impact is … machines going to read your content.”

That observation, from Hamzah Khadim, co-founder of Toronto-based Logik Digital, captures what high-velocity data publishers are now building around.

Generative Engine Optimization Emerges as a Distinct Practice

Logik Digital now centers its work on GEO, generative engine optimization, a discipline distinct from conventional SEO. Where SEO positioned content to rank in a list of links, GEO positions content to be included in an AI’s answer. Khadim describes the shift as moving from keywords to conversations, with the primary goal being part of the AI’s response rather than appearing below a search bar.

The practical work looks different from what most publishers expect. GEO can mean adding large volumes of text to a site’s About Us section, content that human visitors are unlikely to read in full but that AI will process to assess credibility.

“People aren’t gonna necessarily read all of that, but AI definitely will. And it’ll help establish trust.”

For B2B publishers and marketers, the implication is that content strategy now has two audiences with different needs, and the machine audience is becoming the gatekeeper to the human one.

The Economic Fault Line Between Cooperation and Cutoff

The infrastructure cost of AI scraping is not trivial. Lane Becker, president of Wikimedia Enterprise, has described the scale of bot traffic his organization manages: a single bot can click on 50 Wikipedia articles in under a second, and one bot could pull every page of every Wikipedia edition in every language in under half an hour.

“It’s easy for one robot to show up and say, ‘Hey, I want to scoop every single page out of every single Wikipedia [edition] and every single language, and I’m going to do it in under half an hour.’”

Wikimedia’s response has been to formalize the relationship. Wikimedia Enterprise has struck paid, direct, bulk-access deals with Google, Microsoft, Amazon, and Meta, offering a managed channel for AI companies rather than an uncontrolled scraping free-for-all. That model accepts the machine-reader reality and attempts to extract value from it.

Other publishers are moving toward harder lines. USA Today’s CEO stated recently that the traditional Google search business model is dead, that traffic is declining, and that the outlet is approaching a position where it may block Google from the site entirely. According to Wall Street Journal reporting cited by CBC, Reddit and other prominent publishers are weighing similar steps.

Mark Surman, president of Mozilla, has raised the question that underlies the entire debate.

“If those incentives to create original, creative, independent content disappear, does content just come from big and powerful companies, or from governments, or from AI itself?”

The two models, Wikimedia’s collaborative paid partnership and USA Today’s readiness to cut off access, represent the poles of a decision every significant publisher will eventually face. Neither outcome is settled.