Last year, a website visit usually meant a human. Now it probably doesn't.
Cloudflare - which routes traffic for roughly 20% of the websites on the internet - is warning that automated requests from bots and AI agents have overtaken human-generated traffic. The company has visibility into raw web traffic at a scale few organizations can match, making this observation credible.
Traditional bots have always made up a large share of web traffic: search crawlers, content scrapers, uptime monitors, spam operations. What's changed is the second wave - agentic traffic. AI programs that browse the web on behalf of users to complete tasks don't look like old-school bots. They follow links, execute JavaScript, accept cookies, fill in forms. They behave like humans, just faster and in far greater volume. Every person using an AI assistant with web access generates dozens or hundreds of automated site visits per day, often without the destination site realizing it.
Your Analytics Are Probably Wrong
The practical problem for anyone running a website: your traffic data is increasingly unreliable. Most analytics tools filter known bots using user-agent lists and behavioral patterns. But AI agents designed to browse the web often mimic human sessions closely enough to pass those filters. Your reported human visitor count likely includes a growing share of automated requests.
This matters most if you make decisions based on traffic numbers - ad revenue thresholds, feature adoption signals, conversion rates. If a substantial portion of what looks like human engagement is actually agents completing tasks for users, your funnel metrics are off in ways that are hard to detect and harder to quantify.
For subscription or API businesses, the stakes are higher. Rate limits and pricing models built for human usage patterns can be exhausted quickly by a handful of power users running agents continuously against your product.
The Web Built for Humans, Read by Machines
There's a structural issue beneath the traffic numbers. Web pages have always been built for human readers - visual layout, brand copy, persuasion sequences. But an AI agent answering a user's question about your pricing doesn't need your hero image or your testimonials. It needs structured, accurate, machine-readable data.
Sites built for human reading are increasingly less visible to the agents that now sit between you and potential customers. AI assistants regularly surface competitors whose data is cleaner and easier to extract, regardless of which site a human reviewer would prefer.
Cloudflare has been building tools to help: its AI Audit product lets publishers identify which AI crawlers are hitting their sites and set access rules. But the choice isn't simply "block bots" - some of those agents are helping real customers complete purchases, and blocking them means blocking the transaction. The line between a wanted agent and an unwanted one is increasingly hard to draw, and Cloudflare's warning suggests that line is about to become everyone's problem.