For the first time since the web existed, machines generated more of its traffic than people did in 2026. Bots and AI agents accounted for more than 57 percent of web activity, against about 42 percent from humans. The figures come from Cloudflare, which handles traffic and blocks attacks for roughly a fifth of the internet. “Humans will be a rounding error on the internet versus what their agents and their bots are doing,” says Matthew Prince, its chief executive. Behind the shift sits Google’s own product, the AI-written answers at the top of its search results.
Google earned its place at the center of the web through a simple exchange. It runs a program that visits and copies pages across the web, known as a crawler. For two decades, its crawler, Googlebot, sent human readers back to the sites that published those pages. Those visits were how a site earned from its content, until AI answers broke the trade. An AI answer settles a reader’s question on the spot, so the crawler that once brought visitors now keeps them where they searched.
The same crawler feeds those AI answers, and it reaches far more of the web than the crawlers of rival AI companies. Prince says Googlebot reaches 3.2 times more of the web than OpenAI’s crawler and 4.8 times more than Microsoft’s. A wider, fresher pool of pages makes for stronger AI answers, part of why he calls the arrangement an unfair advantage.
That double job is his central complaint. A publisher therefore cannot keep its work out of the AI answers without also vanishing from ordinary search, the traffic source most sites depend on.
Because Google’s AI rewrites the web for its answers without sending readers back, writers have less reason to keep producing the pages those answers depend on.
A fix is being tested. Google has confirmed a new setting that lets a website opt out of AI answers while keeping its normal search rankings. But the setting leaves the single crawler in place, so a publisher has to trust that the switch works as promised. Splitting Googlebot into two crawlers, one for AI answers and one for search, would let publishers block AI use directly.
No such split has happened, and the company running the crawler still handles about 90 percent of how people search the internet.
