Cloudflare: More Than Half of Internet Traffic Is Now Not Human

More than half of Internet traffic is no longer human, according to Cloudflare, which says AI agents have become a second audience for the web. The company is pairing that claim with new controls and payment tools meant to help site owners serve agents and earn money from them.
- Cloudflare says more than half of Internet traffic was not human this year, with daily AI agent requests up more than 1,700% in a year.
- Cloudflare handles an average of 115 million HTTP requests a second, up from 63 million at the end of 2024.
- Separate Search, Agent and Training controls are available on every plan, and Disallow AI Training shipped on September 15.
- Pay Per Use pays publishers when content is actually used, while Monetization Gateway is in closed beta for eligible U.S. customers.
A second audience arrives
Cloudflare says the Internet long had one audience that paid the bills: people who read articles, saw ads and bought subscriptions. Bots existed, but they were mostly large automated operations that did not view ads or pay for anything.
The company's figures point to a sharp change. At the end of 2024, Cloudflare handled an average of 63 million HTTP requests a second. Today that has almost doubled to 115 million, with peaks above 150 million. Over the past year, daily requests from AI agents on its network grew by more than 1,700%, and this year, for the first time, more than half of Internet traffic wasn't human, according to Cloudflare.
Cloudflare describes agents as software acting on behalf of people. They sit between humans and traditional bots: they do not respond to ads, but there is usually a person behind them with a job to get done. For businesses that learn to serve them, the company says agents are additive. For those that don't, they are extractive.
More traffic, less revenue
For thirty years, Cloudflare says, the web ran on one arrangement: sites let search engines crawl them, search engines sent visitors, and the sites turned those visitors into a business. AI has broken that balance. Answer engines read the page and give the reader a summary, which costs a site bandwidth without sending a human to the place where ads or payments happen.
Cloudflare reports that some of the most heavily crawled categories, including Retail, Computer Software, IT & Services, and Financial Services, have seen human traffic decline as much as 40% in less than one year. Every automated request still costs bandwidth, compute and origin capacity, while a growing share carries no referral, ad impression or subscription.
The share of crawler requests made for AI training also rose. In Spring 2025 it was 22%, according to the crawlers' stated purpose. By June 2026 it was 52%. Cloudflare says a blanket "no" is not nuanced enough, because some agent traffic is a customer: an agent booking a table or comparing insurance quotes acts for a real person. Unlike a training crawler, an agent returns each time someone asks about the content, so its traffic grows with the number of questions people ask rather than the amount a site publishes.
Seeing visitors and setting terms
Cloudflare argues a site cannot do business with an audience it cannot see, tell apart, set terms for or charge. Its AI Crawl Control, Business Insights and BotBase tools show who is crawling, what they take and which URLs they want most. With Web Bot Auth, operators including OpenAI, Google and AWS cryptographically sign their agents' requests, so a site can separate a real agent from an impersonator. Cloudflare says it sees more than 500 billion verified bot requests each week.
In July, the company replaced its single "block AI bots" switch with separate Search, Agent and Training controls, available on every plan including Free. Its data showed fewer than 1% of sites on Cloudflare block search crawlers, while 17% block training.
On September 15, it shipped Disallow AI Training, which keeps a site indexed for search while using crawler-specific mechanisms to tell the operator not to use the data for training. Cloudflare says Apple, Google and Microsoft have committed to honor it. New domains now see recommended configurations based on how the site makes money. For ad-supported sites, that can mean disallowing training and blocking agents on pages with ads, since an ad only pays when a person sees it.
Getting paid: Pay Per Use and Monetization Gateway
Cloudflare says the web needs a way to say "yes, if you pay" instead of a binary yes or no. It notes that more than 50 publisher-AI deals have been signed since 2023, but nearly all are bespoke and bilateral between large publishers and large AI companies, so they do not reach most of the web.
Its answer for the rest is Pay Per Use. It does not charge for the crawl but pays when content is actually used. Every buyer is a verified crawler that defines what counts as use and what it will pay. Publishers see the offer, choose whether to opt in and can opt out at any time. The buyer reports each use, Cloudflare checks the reports, bills the buyer and pays the publisher. Publishers can see what was used, when and what they earned, and in some cases which questions surfaced their work.
Monetization Gateway targets cases where accounts, API keys and subscriptions do not fit, such as an agent that wants a single lookup from a service it has never used. In a closed beta, eligible U.S. Cloudflare customers can put a price on anything that passes through the network using the Rules language. When a rule matches, Cloudflare returns an HTTP 402 Payment Required using the open x402 protocol, and the agent pays the seller directly.
What it means for site owners
For site owners, Cloudflare's message is that blocking everything is no longer the only choice. The tools it describes are meant to let a site identify agents, allow search while refusing training, and charge for use. Whether agents turn into a revenue source rather than a cost will depend on how many AI companies take part as buyers, a point the source does not settle.
Cloudflare: The Internet has a second audience


