Automated AI Agents Now Drive Majority of Web Traffic as Machine Economy Takes Hold
Automated artificial intelligence agents and software bots now account for the majority of global internet traffic, according to data from leading cybersecurity and infrastructure firms, marking a structural tipping point in the architecture of the digital economy.
Data released by web infrastructure provider Cloudflare reveals that non-human activity accounted for 57.5% of all web requests in June, surpassing human engagement far earlier than industry forecasts had anticipated. Cloudflare Chief Executive Matthew Prince had previously projected that automated traffic would not cross the 50% threshold until late 2027. Cybersecurity firm Thales reached a similar conclusion in its 2026 evaluation, reporting that non-human activity had reached 53% of overall web volume.
The acceleration has been primarily driven by “agentic” software designed to execute complex tasks online independently, rather than static scrapers or basic automation routines. According to cybersecurity company HUMAN Security, web traffic generated by action-taking autonomous agents—such as those filling out forms or navigating web elements—surged by 7,851% year over year. While automated scrapers grew by 597% over the same timeframe, AI training crawlers, which still account for 67.5% of total AI-driven activity, are steadily declining as a proportion of total non-human web volume.
This demographic shift in web users is upending long-standing digital business models that relied on direct human attention, such as pageview metrics, display ad impressions, and consumer conversion funnels. Because autonomous agents navigate and process digital environments differently than human visitors, technology providers are adjusting their core software architectures.
Payment processor Stripe reports that 70% of requests to its Application Programming Interfaces (APIs)—the core software channels allowing different systems to exchange financial data—are now initiated by AI agents. Financial technology vendor Alpaca recorded a parallel shift, with agent-driven API activity climbing from single-digit percentages in late 2025 to 30% in early 2026. In response, roughly 25% of developers are now prioritizing software agents over human users when designing new interfaces, while more than half cite unauthorized agent traffic as a top security risk.
Major corporate entities including Visa, DoorDash, Coinbase, Ramp, Mercury, ElevenLabs, MoonPay, and Stripe have already launched specialized Command Line Interfaces (CLIs) engineered explicitly to allow automated agents to perform programmatic operations.
However, traditional web monitoring systems struggle to accurately identify and filter this growing agent traffic. Standard security filters often fail to distinguish between human browsing patterns and sophisticated agentic software designed to mimic human interactions. A study conducted by researchers at the University of Bamberg found that legacy bot-detection platforms suffer soft block error rates of 7% to 15%, inadvertently restricting real users while missing sophisticated automated browsers. Digital marketing firm Seer Interactive has similarly alerted clients that undetected agent activity risks corrupting foundational business metrics by skewing session durations and artificially suppressing bounce rates.
While autonomous traffic dominates volume, its direct financial footprint remains constrained by technological limits. Investment research firm PitchBook estimates that the emerging “machine economy” generates approximately $36 billion in annual gross domestic product, representing just 1% of the estimated $20 trillion in economic tasks that could theoretically be delegated to artificial intelligence. According to Pitchbook analyst Rudy Yang, the transition to fully autonomous commercial operations remains blocked by unresolved challenges surrounding automated payment systems, identity verification, and legal liability.
