Cloudflare 首席执行官马修·普林斯告诉 The Verge，在五年内，机器人产生的网络流量可能比人类流量多出 1000 倍。他是在尼莱·帕特尔的 Decoder 播客节目中做出这一预测的。他补充说，他之前关于机器人流量的每一次预测都错了。
Bots could make up 1,000 times more web traffic than humans within five years, Cloudflare chief executive Matthew Prince told The Verge. He made the forecast on Nilay Patel’s Decoder podcast. He added that each of his earlier predictions on bot traffic had been wrong.
之所以会错，是因为机器人的增长速度超出了他的预期。去年 11 月在 Web Summit 大会上，普林斯表示自动化流量将在 2027 年下半年超过人类流量。到了 3 月，Cloudflare 的数据指向 2027 年上半年。5 月，他的团队告诉他，这已经发生了。
They were wrong because bots grew faster than he expected. At Web Summit last November, Prince said automated traffic would pass human traffic in the second half of 2027. By March, Cloudflare’s data pointed to the first half of 2027. In May, his team told him it had already happened.
“五年后，我们认为网上的自动化流量将是人类流量的 1000 倍，”普林斯说。
“Five years from now, we think that automated traffic will be 1,000 times human traffic online,” Prince said.
他说，人类流量将大体保持平稳。增长来自 AI 爬虫和代理。Cloudflare 处于超过 20% 的网站前端。
Human traffic will stay roughly flat, he said. The growth comes from AI scrapers and agents. Cloudflare sits in front of more than 20% of the web.
普林斯说，大约 30 年来，网络的商业模式一直是广告。这种模式对机器人不起作用。> “机器人不会点击广告，”他说。
For about 30 years, the web’s business model has been advertising, Prince said. That model does not work for bots. > “Bots don’t click on ads,” he said.
Cloudflare 正与 Coinbase 和 Stripe 合作，复活 HTTP 402 状态码，这是网络早期的“需要付费”代码。网站所有者可以向 AI 代理收取每次访问页面几分钱的费用。有价值的内容可以定价更高。普林斯说，买方的 AI 公司和卖方的大型发行商双方都愿意。
Cloudflare is working with Coinbase and Stripe to revive HTTP 402, a “payment required” code from the web’s early days. Site owners could charge AI agents a fraction of a penny for each page they access. Valuable content could cost more. Prince said AI companies on the buying side and big publishers on the selling side are both willing.
帕特尔反驳说，这样的市场是否能给人们在网络上发布内容提供理由。雅虎首席执行官吉姆·兰佐内上周表示，微支付永远不足以支撑发行商。
Patel pushed back on whether such a market would give people a reason to publish on the web. Yahoo chief executive Jim Lanzone said last week micropayments will never be enough for publishers.
默认屏蔽 Google 从 9 月 15 日起，Cloudflare 将其免费客户的设置默认配置为屏蔽 Google 使用其网站进行 AI 训练。网站所有者可以关闭该设置。
Blocking Google by default From 15 September, Cloudflare set its free customers to block Google from using their sites for AI training by default. Site owners can switch the setting off.
“如果他们不区分 AI 训练和搜索引擎训练，我们就全面屏蔽他们，”普林斯说。
“If they won’t differentiate between AI training and search engine training, we’re just going to block them across the board,” Prince said.
他说，谷歌已承诺在其爬虫方面提供更多透明度。普林斯表示，在英国的一项裁决后，谷歌暗示其爬虫会说明其在每个页面上的行为。发布商届时可以选择，谷歌将把这一做法推广到英国以外。布鲁塞尔也在询问发布商，谷歌的 AI 退出选择机制是否有效。
He said Google has committed to more transparency about its crawler. After a ruling in the UK, Google signalled that its crawler would say what it is doing on each page, Prince said. Publishers could then choose, and Google would apply this beyond the UK. Brussels is also asking publishers whether Google’s AI opt-out works.
普林斯和他的妻子拥有《帕克纪事报》，这是一份位于犹他州帕克城的地方报纸。他预计今年该报从 AI 许可协议中获得的收入将超过数字广告收入。他说，在 AI 许可方面，Reddit 每个 token 的收入是《纽约时报》的 7 到 14 倍。
Prince and his wife own the Park Record, a local newspaper in Park City, Utah. He expects it to make more from AI licensing deals than from digital advertising this year. In AI licensing, Reddit earns seven to 14 times more per token than The New York Times, he said.