Humans won't disappear from the internet. But from now on, we are no longer the only paying users.

Article author and source: AI New Era

Will the internet no longer belong to humans?

In five years, humans will become an "error" on the internet.

During the Q2 2026 earnings call, Cloudflare's CFO Thomas Seifert delivered this blunt statement to investors.

Who is Cloudflare?

More than one-fifth of websites worldwide route their traffic through this vast network—and over half of this traffic is no longer generated by humans.

This is the first time machine traffic has surpassed human traffic in Cloudflare's observations.

More interesting is the time.

Cloudflare CEO Matthew Prince originally predicted the threshold would be crossed by the end of 2027, later revising it to early 2027. In May 2026, the team came to him and said: “It’s already happened.”

Faster than I expected.

This breach occurred more than a year earlier than Prince’s own prediction. The statement from Seifert that “humans will become an error” is also derived from this trend:

If this growth continues, in five years, machine traffic will be 1,000 times that of human traffic.

This means that the websites you browse daily, the links you click, and the username you argue with in the comments section are likely no longer human.

Diluted by a thousandfold machine traffic, your digital footprint on the internet shrinks to a mere "error."

You think you're browsing the internet, but it's actually the machine doing it.

This extra machine traffic didn’t appear out of nowhere—much of it was generated by the AI you invoked.

Prince once made a particularly vivid analogy.

You might open five websites to compare prices when buying a digital camera, but if you leave it to AI, it’ll browse 5,000 for you.

A single human thought is amplified into thousands of machine actions.

This is the real reason for the explosive traffic growth: it’s not because hundreds of billions of new internet users have come online, but because behind every person now stands an tireless army of intelligent agents.

Therefore, the real change is that going online is shifting from "you going yourself" to "sending a machine instead."

Pitchbook breaks down the agent's "fan-out effect": on the left, a human makes a few linear calls; on the right, a single command is decomposed into planning, tool invocation, parallel sub-agents, and aggregation, resulting in dozens or even hundreds of machine actions.

For the past thirty years, web pages have been designed for humans—typography needed to look good, interactions needed to be intuitive, and ads needed to keep you engaged for just three more seconds.

But now, an increasing number of "visitors" have no eyes at all—they don’t pay attention to layout or design, grab the data, and leave.

The rules established for humans have failed for the first time in the face of machines.

The Internet Death Theory has gone from a joke to reality.

Regarding machine traffic, this has long been predicted online as the "Death of the Internet" theory.

It means: What you think is a vibrant internet is mostly machine-generated content, with most interactions being bot-driven—real humans have long become the minority.

In Gibraltar, machines account for 93% of web requests, leaving humans at just 7%. The "dead internet" is no longer a prophecy in some places.

This conspiracy theory, mocked for years, has been repeatedly revived in the AI era and has now made its way into earnings calls, becoming a number presented to investors.

You’ve likely already noticed this: endless AI-generated images and text, comment section accounts with unnatural phrasing that sound copied and pasted, and sudden spikes in likes and shares overnight.

Cloudflare announced on the call: More than 50% of traffic is not human.

But when you review its public data page, Radar, the caption for the "Machine vs. Human" chart is clear: it only counts HTTP requests that return HTML content.

In other words, only count the act of "web requests" themselves.附属 requests such as images, scripts, and fonts are not included, nor is machine-to-machine traffic like APIs.

The determination is based on a Bot Score assigned to each request: scores of 1 to 29 indicate "likely a bot," while scores above 30 indicate "likely a human."

So a more accurate statement is: On the Cloudflare network, in HTTP requests returning HTML, machines account for more than half according to their own scoring criteria.

Radar "Machine vs Human" real-time chart: under this metric, machines account for 60.4%, and humans for 39.6%.

And even this half of the traffic isn't all AI—it also includes search engine crawlers, website monitoring tools, security scanners, and malicious scripts, a whole bunch of "old familiar faces."

But the most explosive and game-changing growth is undoubtedly in AI.

According to HUMAN Security’s 2026 report, AI-driven traffic still consists of approximately 67.5% training crawlers, though their share is declining; real-time scraping crawlers increased by 597% over the year; and AI agents capable of independently clicking links, filling out forms, and completing purchases surged by 7,851% in one year.

In observed AI bot traffic, OpenAI accounts for approximately 69%, Meta for 16%, and Anthropic for 11%. (Source: HUMAN Security 2026 State of AI Traffic Report)

The observed bot traffic is alarmingly concentrated.

OpenAI accounts for approximately 69%, Meta about 16%, Anthropic around 11%, and the remaining dozens of companies together make up less than 5%.

In other words, three AI companies are reading the internet on everyone’s behalf.

Prince described this as a platform shift on the scale of "desktop to mobile."

The only difference is speed: It took ten years to move from PCs to smartphones back then; this time, it will be over in just a few months.

The internet is beginning to be renovated.

When the visitors change, the internet must be renovated to suit the new visitors' tastes.

This is already happening.

An increasing number of websites are beginning to offer content formats specifically designed for AI, converting visually embellished pages meant for humans into plain text that machines can easily process.

The art of search engine optimization has quietly changed: it used to be about "how to get people to find me," but now it's about "how to get AI's answers to cite me."

When the intended audience for content shifts from human readers to "models," much of what you see online isn't written for you—it's written for the AI working on your behalf.

You merely glance at the conclusions it has organized. From now on, there is a machine intermediary between you and the information.

The old model of the internet is coming to an end.

Here, there's still a more practical issue lurking.

For the past two decades, the internet has been free to use, thanks to an unspoken exchange:

You view ads; the website earns revenue from them and uses that to provide content to you for free.

This exchange operates under a default assumption: the people viewing the content are human. Humans view ads, and humans make impulsive purchases.

Machines don’t watch ads, and they certainly don’t impulsively buy something just because they’ve been longing for it. When most visitors are machines, the foundation of the advertising business begins to crumble.

Prince bluntly stated: The old internet model, defined by Google and sustained by advertising for nearly three decades, is coming to an end.

The new version he provided is: machine-to-machine, pay-per-use. A tiny toll, less than a fraction of a cent, is charged for each AI request.

In fact, Cloudflare has already begun fixing this toll booth and has unveiled the entire system that enables AI to pay upon request.

Prince even claimed he wants to build a payment network three orders of magnitude larger than Visa, specifically designed for machines.

We spent thirty years moving the entire world onto the internet. Then, in less than three years, we turned it into a place primarily for machines.

So, what is the foundation of this new internet?

The investment bank PitchBook introduced a term: the machine economy.

How large is the current economy? Startup Forsy estimates that the global "agent GDP" is currently around $36 billion per year.

Sounds significant, but only about 1% of tasks AI can actually take over are currently being run through agents.

The money side has only just processed 1%. On the traffic side, millions of requests per second pass through, with almost none resulting in a payment.

Although the market cap is still small, the slope is already very steep.

Data from the payment company Stripe shows that 70% of command-line requests now come from agents.

The last internet was sustained by people viewing ads; the next one might be funded by machines paying tolls.

Humans won't disappear from the internet. But from now on, we are no longer the only paying users.

And network traffic must also be redefined.

Reference materials:

• https://radar.cloudflare.com/
• https://x.com/eastdakota/status/2062212701414187452
• https://www.humansecurity.com/learn/resources/2026-state-of-ai-traffic-cyberthreat-benchmarks/
• https://pitchbook.brightspotcdn.com/9e/4f/cbd3ad0d49e3aba868c3f07f0d19/q3-2026-machine-economy-rising-the-market-of-limitless-cognition.pdf