Real eyes. Real attention. Real intent to buy, share, or act.

That assumption has now been overturned.

According to Cloudflare's Radar data - one of the most comprehensive traffic monitoring systems in the world - automated bots now account for 57.5% of all HTTP requests to web content. Human generated traffic has fallen to 42.5%. For the first time in the internet's history, machines are generating more web requests than people.

The CEO of Cloudflare, whose network processes a significant share of global internet traffic, acknowledged the milestone publicly, admitting that even he was caught off guard. He had predicted the crossover would occur by the end of 2027. It arrived roughly 18 months early.

The shift in internet traffic

Bots have existed since the earliest days of the web. Google's search crawlers, spam engines, fraud systems, and monitoring tools have long operated in the background. But those were simple, repetitive programs. The new wave is categorically different.

Today's bots are artificial intelligence (AI) agents: software that browses the internet on behalf of human users, performing complex, multistep tasks that once required sustained human attention. When a person asks an AI assistant to research a product, plan a trip, or summarise recent news, the assistant does not retrieve a cached answer. It dispatches agents across the web, visiting dozens, hundreds, or even thousands of pages in seconds to assemble a response.

Consider the arithmetic. A person shopping for a smartphone might visit five or six websites. An AI agent executing the same task might visit five hundred. One human intention now generates thousands of machine actions. Multiply that across hundreds of millions of users now routinely querying AI assistants, and the shift in internet traffic becomes not just plausible but inevitable.

The internet is not emptier. It is, in raw transactional terms, busier than ever. But the activity is increasingly mechanical.

Why every LLM query pushes humans further back

There is an irony at the heart of the AI assistant revolution that most users never consider. Every time a person delegates a search to ChatGPT, Gemini, Perplexity, or any comparable tool, they are not simply getting a faster answer. They are handing over their role as an active internet participant to a machine.

The human types a question. The AI agent does the browsing, the reading, the comparison, the filtering, and the synthesis. The human receives a summary. In this transaction, the person has moved from being a direct actor on the web to being a passive recipient of machine-curated conclusions.

At scale, this matters enormously. The web was built around human attention : the signals it generates, the choices it makes, the paths it follows. When those signals are replaced by agent-generated requests, the data that powers everything from search rankings to content recommendations to advertising targeting becomes structurally distorted. The web increasingly reflects what machines ask for, not what humans want.

The crisis coming for Google and for every website

The commercial internet was engineered around a single economic mechanism: human attention converted into advertising revenue. Cost-per-click, cost-per-impression, conversion funnels, affiliate links, programmatic advertising. Every component of this system assumes that a human is viewing a page, noticing an advertisement, and potentially acting on it.

AI agents are indifferent to all of this. They do not see banner advertisements. They do not linger on lifestyle content. They are not susceptible to impulse purchases triggered by clever placement. They extract the information they need and move on, leaving no revenue trail behind them.

This creates a paradox that should alarm every publisher, blogger, and digital business: a website may now receive more traffic than ever before and earn less money from it. The busiest internet in history may simultaneously be the least monetisable.

Google, whose advertising empire depends on human users clicking search results and viewing ads on partner websites, faces a structural challenge that goes beyond competition from AI assistants. If the majority of web traffic is generated by agents that bypass the entire advertising layer, the foundation of the open web's funding model begins to erode. Revenue may not collapse immediately, but the trajectory is difficult to ignore.

What survives, and what must be reinvented

It would be a mistake to conclude that the internet is simply dying. The distinction that matters is between the technical traffic layer and the human engagement layer, and they are separating in real time.

Humans still dominate in time spent online. Streaming platforms, social media feeds, messaging applications, and entertainment services generate sustained human engagement that no bot replicates. The emotional and social dimensions of the web remain irreducibly human. A conversation, a shared video, a moment of online community and these cannot be automated.

But the infrastructure of the web: the page requests, the data fetching, the indexing, the comparison shopping is shifting decisively toward machines. Cloudflare's own country-level data illustrates just how far this has advanced in some parts of the world, with bot traffic shares exceeding 90% in certain jurisdictions, largely driven by data centre concentration.

For website operators, the strategic question has already changed. Search engine optimisation, the discipline that governed digital visibility for two decades, was built around one question: will a human click my link? The new discipline, already being called generative engine optimisation, asks a different question entirely: will an AI agent read my page, trust its content, and recommend it to a user?

These are not the same question. The first rewards visual design, emotional resonance, and keyword density. The second rewards clarity, factual accuracy, structured data, and machine readable authority signals. Websites designed for human eyes may be invisible to the agents that increasingly mediate access to information.

A turning point, not an ending

The crossing of the bot to human traffic threshold is not the death of the internet. It is the end of one internet and the uncertain beginning of another.

The first internet: human first, attention driven, and commercially sustained by advertising flourished for thirty years. It democratised information, created industries, and connected billions. Its foundational assumption, that a person is on the other end of every request, served it well.

That assumption is no longer true. And almost everything built on top of it , from advertising models to content strategies to platform economics, will need to be rethought.

The transition will be uneven, disruptive, and in many respects uncomfortable. But it is already underway. The data does not lie, and the Cloudflare charts tell a story that is, at this point, beyond debate.

The question is not whether the internet is changing. The question is whether the people, businesses, and institutions that depend on it are changing fast enough to keep up.

Pravin Kaushal is director-Mrikal (AI/Data Center) and a young alumni member, Government Liaison Task Force, IIT Kharagpur. He posts on X as @ipravinkaushal