Cloudflare is tightening default access to the AI reptile and focusing on publisher content fees. The company indicated that, as of 15 September 2026, any “mixed-use” reptiles that capture content on advertising pages and are used for search, AI agency and model training will be stopped by default unless the site owner changes the set-up itself.
Default intercept on 15 September
This adjustment will be applied to new Cloudflare clients, new sites for existing clients and all existing free users. In other words, it will be more difficult for such reptiles to continue to capture content in the same capacity in the future if the site does not proactively release access.
The judgement of Cloudflare is that many websites are willing to continue to be captured by search engines and to be found in selected AI services, but do not want content to be used directly for training or proxy services without compensation.
Search requested to be separated from AI.
In the note, Cloudflare mentions that the largest search engine in the world has about twice the amount of information available to other AI companies, and suggests that such platforms bind search to AI uses, making it difficult for the site to avoid AI use while maintaining search exposure.
Google has in the past refuted similar claims. The public position is that the website can choose not to allow content to be used for training and for AI products such as Gemini Apps, Vertex API, without prejudice to Google Seach. However, Googlebot itself captures content for search, and AI Overviews and AI Mode of search results rely on this system.
Fee model extended to usage
In addition to access control, Cloudflare is promoting a content-fee tool. Pay Per Crawl has been extended to Pay Per Use, which was previously introduced by the company for AI robots.
According to Cloudflare, the new model is not only a charge when content is captured, but a fee when content actually produces value. The first partners were Ceramic.ai and You.com. When the publisher chooses to join, the publisher receives income if the content appears in the Ceramic AI search results, or if you.com calls its paid content.
Cloudflare also states that this change has also helped to reduce the bandwidth and computing consumption of publishers. The company data show that more than half of the AI reptile flows are spent on duplicate capture of unchanged pages. For websites that rely on advertising and subscription income, this means that content is read over and over, but not necessarily in return.