Website traffic from automated bots now accounts for more than 99 percent of server requests hitting some online businesses, with the economics falling entirely on the content provider.
Nick Gray, founder of the Austin, Texas-based philanthropy database PatronView, published findings Aug. 7, 2026, showing 214 non-human page loads for every human page load recorded by his analytics.
The findings detail a year-long battle against automated traffic, with one Anthropic AI search crawler reading PatronView's pages 35,000 times for every human visitor it referred.
Gray's analysis draws a distinction between server logs and JavaScript-based visitor analytics tools like Plausible or Google Analytics. While his server answered 2.5 million requests and served 1.28 million full pages in the week his report was published, Plausible recorded only 5,977 human pageviews—bots generally do not execute JavaScript.
PatronView is a 1.5 million-page research database covering American museum and cultural institution donors. It compiles data from public philanthropic disclosure documents, including IRS 990 forms, donor walls and annual reports. Gray said his own periodic data collection differs from the constant, high-volume automated requests his site now receives daily.
Other AI crawlers add to the traffic burden. Amazon's AI crawler read PatronView pages 117,000 times in a single day at one point without referring any human visitors. Other operators, including Arvid Kahl of Podscan, said they dedicate significant operational work to keeping servers from being overwhelmed by bot traffic.
Before Gray isolated individual AI crawlers, the site experienced traffic spikes tied to geography. In November 2025, roughly 4,000 sessions appeared over a few days, each visiting exactly one page with a 99 percent bounce rate and no referrer. Those sessions concentrated on fund pages that typically draw only 10 percent of real visitor traffic.
A larger incident hit April 22, 2026, when PatronView recorded 3.6 million requests in a single day from 361,844 unique IP addresses, the vast majority located in China.
Cloudflare's Managed Challenge system absorbed 1.18 million of those requests within the first 10 hours. A notable share passed the challenge, indicating sophisticated bot activity.
Gray blocked the entire country of China at the network edge on April 23. He later extended the same block to Vietnam and Singapore after observing similar automated traffic patterns from those regions.
The operational costs are becoming a hard-dollar problem for smaller operators. PatronView incurs roughly $90 per month in costs directly tied to bot traffic. Some web developers say large language model crawlers from companies such as Meta can generate more than 1,000 requests per second from single server clusters.
For now, the financial burden of AI training infrastructure falls on content creators rather than the companies consuming that content—and there is no mechanism forcing the latter to compensate the former.
