The CEO of Cloudflare says that people do not check the source links of IA chatbots

Companies that develop a generative AI always make a duty to say that they include links to websites in the responses that their chatbots generate for users. But Cloudflare CEO Matthew Prince revealed to Axios that the research traffic references continue to fall. Publishers are faced with an existential threat, he said, because people do not click on these chatbot links and count more and more on AI summaries without digging more deeply.

Prince said Axios 10 years ago, Google sent an editor a visitor for the two pages he had crawled. Six months ago, the ratio was a visitor for all six pages, and now it is one for 18. Openai sent a visitor to a publisher for each 250 pages that he crawled six months ago, while Anthropic sent a visitor for 6,000 pages. These days, Openai sends a visitor to an editor for each 1,500 pages, while Anthropic sends a visitor for 60,000 pages.

People have come to trust the chatbots in recent months. The problem for publishers is that they do not earn advertisements if people do not click on links leading to their websites, and that is why Prince encourages them to take measures to ensure that they are paid enough. Prince said Cloudflare is currently working on a tool to block robots that scratch content for large languages ​​even if a web page already has a “crawl” instruction. If you remember, several points of sale had pointed out in 2024 that IA companies had been Ignore the website exclusion protocol for websitesor robots.txt, files and taking their content anyway to train their technologies.

Cloudflare looked for means To block scrapers Since last year. But it was not until March that Cloudflare officially presented Ai LabyrinthWho uses the content generated by AI to “slow down, confuse and waste the resources of AI robots and other robots that do not respect the” crawl “guidelines.” It works by connecting an unauthorized robot a series of pages generated by AI which are sufficiently convincing but which do not really have the content of the site the tool it protected. In this way, the crawler ends up wasting time and resources.

“I go to war every day with the Chinese government, the Russian government, the Iranians, the North Koreans, probably the Americans, the Israelis, all who try to hack our customer sites,” said Prince. “And you tell me, I can’t stop a Nerd with a company C in Palo Alto?”