Bots now outnumber humans online: what it means for your screen time

The data: what Cloudflare measured

During its Q2 2026 earnings call, Cloudflare revealed a milestone it had expected to reach in 2027 but hit two years early: in May 2026, bots surpassed humans in web traffic flowing through their network.

The current split is 60.5% non-human traffic across all HTTP requests to HTML content. Cloudflare's CFO, Thomas Seifert, went further in his projection: within five years, non-human traffic will be 1,000 times greater than human traffic. His exact words: "humans will be a rounding error on the internet."

There's an important nuance to hold onto: human traffic is not declining. It's growing. But non-human traffic is growing at a pace that makes it statistically irrelevant by comparison. It's not that people will use the internet less. It's that the internet will be filled with machines talking to each other at a scale that eclipses everything else.

According to the data published by Cloudflare on their blog (August 2026), the company no longer merely measures this phenomenon: it has turned it into a tool. More on that below.

What this means for you

At DoctorScroll we talk about the attention economy as the system competing for your minutes. That idea just became more concrete. Human attention is no longer the most abundant resource in the digital medium: it's the scarcest.

Non-human traffic includes crawlers that index content, bots that generate entire pages, agents that consume APIs, scrapers that extract data, and automated systems that post, reply, and react without any human in the loop. While you try to spend less time in front of a screen, billions of machine-generated requests fill the space you leave behind.

This has a direct consequence for how you inform and connect with others online. As we explain in how the algorithm shapes your invisible profile, your digital profile is already built by systems that don't necessarily show you what's best for you — they show you what keeps you engaged. When the medium is bot-dominated, telling apart what's human from what's generated becomes a digital wellbeing skill, not just a technical one.

The infrastructure measuring this: Radar Researcher

Cloudflare didn't just publish the number. During its Agents Week on August 7, 2026, it launched Radar Researcher, an AI tool that lets you ask questions in natural language about global internet traffic, network outages, and connection quality, with real interactive charts.

The tool is built on top of Cloudflare Radar, the public open-data platform that has existed since 2020 and offers a free API. Its architecture is a showcase of the agentic era: a Worker combined with the Agents SDK, conversations stored as Durable Objects with SQLite, open models (Kimi K2.7) with a fallback chain of three model families, and everything routed through an AI Gateway.

One detail that matters: users can audit the model's reasoning and the data calls it made to answer. At a time when trust in AI is a central concern, a tool that shows you where every figure came from is rare and valuable. Its intended audience — journalists, researchers, and network engineers — needs exactly that.

Who defends the human side

Cloudflare has also deployed tools to fight unauthorized bots. The precursor is a continuous session-wide behavior evaluation system: 206 million events evaluated across more than 73,000 zones in 24 hours. On top of that sit three layers: Adaptive Intelligence, a machine-learning engine that auto-updates as bots change their patterns; AI Labyrinth, which generates fake AI pages to trap unauthorized crawlers; and a "poison" mode that deliberately feeds incorrect data to systems scraping content without permission.

In other words: an automated war is already underway between bots trying to get in and systems trying to detect them. Legitimate human traffic is increasingly a protected minority in the middle of that battlefield.

What you can do

You can't control what percentage of internet traffic is bots. But you can control how much of your attention feeds that ecosystem. Four concrete steps:

• Set time limits on your most-used apps. It's not willpower — it's friction. The extra seconds it takes to dismiss the limit are enough to decide whether you actually want to keep going.
• Reduce generated noise. Much of what reaches your feed is already generated, recycled, or amplified by automated systems. Filtering sources, following real people, and limiting time in algorithmic feeds directly reduces your exposure.
• Reclaim slow attention. The first step to not competing with machines for your own time is to practice activities that don't compete: reading, walking, talking without a phone nearby, writing. Anything that doesn't need a scroll or a notification.
• Track your screen time and review it weekly. The how to reduce screen time guide has concrete techniques, from grayscale mode to where you physically place your phone.

The bottom line

The 60.5% figure is not an apocalypse. It's a snapshot of the landscape you're already in. The internet is not going back to being a mostly human medium. What can go back to being mostly yours is your time.

If non-human traffic is on track to be 1,000 times greater than human traffic in five years, the question isn't whether the internet will fill up with bots. It already has. The question is how much of your attention you'll keep handing over to a medium that is increasingly designed to capture it automatically — and how much you'll reserve for what actually depends on you.