Matthew Prince, CEO of Cloudflare, confirmed what many suspected: bots already generate 57.5% of page requests on the network, compared to 42.5% of people.
Prince had predicted that crossover for the end of 2027, but artificial intelligence agents brought the date forward eighteen months.
For the first time since Tim Berners-Lee envisioned the web, humans are browsing as minority visitors in their own creation.
What does it mean to inhabit a network where most readers are machines?
The day the network changed inhabitants
By: Gabriel E. Levy B.
When Tim Berners-Lee proposed the World Wide Web in 1989, he described it as a space for people to share knowledge with each other. Everything else was built on that premise: the forms, the passwords, the CAPTCHAs that ask if you are a robot, the advertising that finances the free content.
For thirty-five years, the architecture of the internet assumed that someone was breathing on the other side of the screen.
That era is over. Cloudflare’s measurement shows that machines are already requesting more pages than people.
The data confirms a trend that Imperva had been documenting: in 2024, automated traffic reached 51% of the total and surpassed human traffic for the first time in a decade; in 2025 it climbed to 53%.
What is revealing is not in the figure but in the cause.
Bot traffic grew for years at a predictable rate, driven by search engine trackers and cybercrime. The recent leap has another origin: artificial intelligence agents, programs that browse, read, extract and act on the web on behalf of the models that created them or the users who ship them.
The new inhabitants of the web
An AI agent doesn’t visit a page like a person does. It runs through it in milliseconds and extracts what it needs. Then jump to the next one. Some collect content to train large language models. Others are looking for real-time information to answer the question someone asked a chatbot. The most recent ones execute complete tasks: they compare prices, fill out forms, schedule appointments and buy tickets.
Cloudflare’s figures measure the stampede. GPTBot, OpenAI’s tracker, went from accounting for 4.7% of verified bot traffic in July 2024 to 11.7% a year later. Anthropic’s ClaudeBot rose from 6% to about 10% in the same period.
ByteSpider, the tracker of ByteDance, the parent company of TikTok, is among the most voracious on the planet. Prince attributed the advance of his forecast to the rise of “agentic” AI, the wave of autonomous assistants that Big Tech launched between 2025 and 2026.
The phenomenon feeds on itself. Each new model needs more data to train and each agent needs to navigate to fulfill its task. Machines read for other machines to write, and what they write goes back to the web, where other machines will read it.
An economy that runs out of clicks
The underlying problem is economic. The open web works on an unspoken pact: search engines crawl publishers’ content and, in return, send them visitors who see ads or pay for subscriptions.
AI agents broke that pact. They crawl thousands of pages and return next to nothing.
Cloudflare measured that some AI scrapers pull tens of thousands of pages for every visit they refer.
The effects are already measured
An Ahrefs study of 300,000 keywords found that the presence of an AI-generated summary on Google reduces clicks to the first organic result by 58%.
The Pew Research Center analyzed 68,879 real-world searches and found that when an AI summary appears, only 8% of users click on a link, compared to 15% when it doesn’t. In the news sector, searches that end without any clicks jumped from 56% to 69% in a single year, according to Similarweb.
The industry’s response is just beginning. Cloudflare launched “Pay Per Crawl,” a system that allows publishers to charge AI agents for access to their content, under a simple logic: if machines are going to be the main readers, let them pay for admission. Prince sums it up with calculated optimism: If bots fund the content, the web can still be free for humans.
The shadow that does not disappear
It is advisable not to idealize the new landscape. A dark streak persists within automated traffic: Imperva estimates that malicious bots, dedicated to fraud, credential theft, inventory hoarding, and user impersonation, account for 40% of all network traffic, and that the AI-powered attacks that the firm mitigates daily went from 2 million to 25 million in a year. The World Federation of Advertisers estimates that ad fraud exceeded $50 billion by 2025. Latin America contributes 15.2% of the world’s malicious bot transactions and Brazil ranks second among the most attacked countries, only behind the United States.
But even that data reinforces the central thesis: most of the recent growth has come not from crime but from legitimate automation. The web is filling up with machines that work, not just machines that steal.
The case of the theory that ceased to be a conspiracy
In 2021, journalist Kaitlyn Tiffany documented in The Atlantic an idea circulating in fringe forums: the “dead internet theory,” according to which organic human activity would have been replaced by bots and automated content. For years, the tech community treated it as digital folklore, an entertaining paranoia with no statistical support.
The measurements of 2026 turned it into something else. The internet isn’t dead, but it’s no longer mostly human, and that distinction matters less than it seems. When you publish an article, the probability that your first reader will be a machine exceeds that of a person. When a brand buys advertising, a good part of the impressions are generated by software. When a search engine responds, more and more often no one visits the source.
Prince rejects the apocalyptic version, arguing that AI allows more people to create and post than ever before. He may be right. The question that remains open is one of institutional design: who finances knowledge in a network where the majority of readers do not see ads, do not pay subscriptions, do not share links and do not feel gratitude.
In short, humans lost most of the internet and artificial intelligence agents explain the outcome. Cloudflare puts automated traffic at 57.5%, and AI trackers doubled their share in just one year. The pact that sustained the open web, tracking in exchange for visits, is no longer fulfilled, and models such as the collection for tracking barely rehearse a way out. The network that was designed to connect people today connects, above all, machines with each other. Deciding whether these machines will pay for the knowledge they consume will define who can continue to produce it.
References
AdExchanger. (2025). The AI search reckoning is dismantling open web traffic, and publishers may never recover. https://www.adexchanger.com/publishers/the-ai-search-reckoning-is-dismantling-open-web-traffic-and-publishers-may-never-recover/
Ahrefs. (2025). AI Overviews reduce clicks [December 2025 update]. https://ahrefs.com/blog/ai-overviews-reduce-clicks/
Berners-Lee, T. (1989). Information management: A proposal. CERN. https://www.w3.org/History/1989/proposal.html
Cloudflare. (2025). From Googlebot to GPTBot: Who’s crawling your site in 2025. https://blog.cloudflare.com/from-googlebot-to-gptbot-whos-crawling-your-site-in-2025/
Imperva (Thales). (2025). 2025 Bad Bot Report: The rapid rise of bots and the unseen risk for business. https://www.imperva.com/blog/2025-imperva-bad-bot-report-how-ai-is-supercharging-the-bot-threat/
Imperva (Thales). (2026). Bad Bot Report 2026: Bad bots in the agentic age. https://www.imperva.com/blog/bad-bot-report-2026-bots-agentic-age/
NBC News. (2026, June 3). Bot web traffic has overtaken human web traffic, data shows. https://www.nbcnews.com/tech/tech-news/bot-web-traffic-overtaken-human-web-traffic-data-shows-rcna348522
Pew Research Center. (2025, July 22). Google users are less likely to click on links when an AI summary appears in the results. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
PPC Land. (2025). Bot traffic costing advertisers billions as fraud detection fails, investigation reveals. https://ppc.land/bot-traffic-costing-advertisers-billions-as-fraud-detection-fails-investigation-reveals/
Radware. (2026). 2026 Global Threat Analysis Report. https://www.globenewswire.com/news-release/2026/02/19/3240861/8980/en/Radware-2026-Global-Threat-Report.html
Tiffany, K. (2021, August 31). Maybe you missed it, but the internet ‘died’ five years ago. The Atlantic. https://www.theatlantic.com/technology/archive/2021/08/dead-internet-theory-wrong-but-feels-true/619937/
