The internet and the foundational web browsers designed to open it were originally engineered from the ground up for human users. However, as automated systems and autonomous software take over an increasingly massive share of daily web traffic, the digital landscape is undergoing a profound architectural shift. Acknowledging this reality, Cloudflare has launched Kitesurf, an innovative new web browser built exclusively for artificial intelligence agents rather than human beings.
Kitesurf operates as a cloud-hosted browser tailored specifically for the software programs that routinely check prices, compare products, evaluate specifications, and reorder goods on behalf of human shoppers. Currently, the browser is available for free during its beta phase inside Browser Run, which serves as Cloudflare’s dedicated browser automation service.
This infrastructure launch arrives at a pivotal moment in the evolution of internet traffic. Automated systems now account for a staggering 57.4% of all web requests worldwide, a figure that climbs even higher to 68.6% across North America, according to data from Cloudflare cited by PYMNTS. This major milestone—where non-human traffic officially eclipsed human browsing—arrived sooner than industry leaders originally anticipated. Cloudflare CEO Matthew Prince noted in June that agent-driven traffic grew at an exponential pace, outstripping forecasts that did not expect automated systems to surpass real people until the following year.
Despite this massive volume of automated activity, these sophisticated AI agents have largely been forced to run on traditional browsers originally designed for human convenience. Open-source engines like Google’s Chromium—which powers popular consumer applications like Chrome—carry a heavy burden of memory and computing overhead. According to Cloudflare’s official launch post, running one traditional browser instance per individual AI agent makes the operational scale far too expensive. Features such as customizable visual themes, multiple tabs, and browser extensions provide zero functional utility to an AI agent, yet they rack up substantial token costs and scalability bottlenecks. In contrast, Kitesurf is designed to launch a dedicated session for a single request and immediately discard it afterward, optimizing resources for machine-to-machine interactions.
The introduction of Kitesurf also positions Cloudflare among a growing number of technology companies racing to develop next-generation web browsers. Nevertheless, Kitesurf is explicitly not pitched to consumers as an alternative to Chrome or other human-centric interfaces. The broader ecosystem of consumer-facing AI browsers has already begun to spark friction with major digital retailers. For instance, consumer-facing AI developments have previously collided with commercial giants, highlighted by Amazon’s legal action against Perplexity over its Comet browser.
Kitesurf Uses 3 to 7 Times Less Processing Power
To demonstrate the efficiency of the new system, company benchmarks compared Kitesurf directly against traditional Chromium performance across 14 distinct webpages, taking the median measurements from five separate runs. The results underscored significant resource savings. Capturing a screenshot required 3.1 times less processing power and consumed 4.7 times less memory when executed through Kitesurf. Furthermore, extracting plain text data off a target webpage demanded 3.8 times less processing power and used an impressive 7 times less memory than its legacy counterpart.
These efficiency gains, however, come with a specific operational trade-off involving raw execution speed. According to the company’s performance data, rendering screenshots finished about 1.8 times slower on Kitesurf compared to Chromium, while text extraction operations ran approximately 1.7 times slower.
At its current stage of development, Kitesurf cannot yet handle every complex task required by a comprehensive shopping agent. Cloudflare has acknowledged that the browser is not yet equipped to play streaming video, maintain long-term account login states across multiple sessions, or successfully navigate certain advanced bot-detection and security challenges deployed by websites. For those heavier operational requirements, Chromium-based tools still handle the workload. Looking ahead, Cloudflare has outlined plans to open-source Kitesurf, enabling enterprise customers and developers to deploy and manage their own customized copies.
Merchants Weigh Access as 132 Million Shop With AI
While more efficient browser technology lowers the operational cost of running automated software, it does not automatically resolve the broader commercial question of digital access and merchant permissions. E-commerce platforms are actively wrestling with how to handle the influx of automated buyers. In September, Amazon blocked Meta’s Muse AI agent from accessing its marketplace, following earlier aggressive efforts by the retail giant to bar shopping agents deployed by Google and OpenAI. Conversely, Shopify moved in the opposite direction just a day later, integrating its Shop Pay checkout solution directly into Meta’s Muse agent.
Cloudflare itself operates across both sides of this commercial divide. The company recently introduced new administrative controls designed to help ad-supported websites manage automated traffic, including a setting that blocks user-directed AI agents on any page displaying advertisements.
Merchants are increasingly forced to make these high-stakes access decisions with remarkably limited visibility into consumer behavior. Nearly 132 million adults in the United States have already purchased a retail product with the assistance of artificial intelligence, according to a September PYMNTS Intelligence report titled "Will the 2026 Shopping Season Go Agentic?"
Despite this widespread consumer adoption, merchant preparedness remains low. Only 23% of merchants are currently able to accurately and clearly identify both AI-driven traffic and the specific consumer purchases that follow, as revealed by the "Global Digital Shopping Index: Merchant Edition." That comprehensive study, published by PYMNTS Intelligence in partnership with Visa Acceptance Solutions, drew on a March survey of 1,185 retail merchants distributed across the United States, Brazil, and the United Arab Emirates.
Additional industry research highlights the growing urgency of the trend. Ninety percent of merchants already report seeing tangible signs of AI shopping activity on their platforms, according to an October report titled "The Agent Advantage: How Merchants Plan to Win AI Shopping," which surveyed 60 senior e-commerce decision-makers. However, the data underscores a persistent tracking gap: precisely none of the merchants selling goods through their own proprietary websites or mobile applications are currently able to consistently track an AI-driven visit all the way through to a completed purchase.
