Cloudflare said on 4 August that it has removed Perplexity AI from its verified-bots programme and is now blocking the company’s traffic after detecting what it calls “stealth crawling” across tens of thousands of websites. According to a technical blog post, Perplexity’s declared bots generate about 20–25 million requests a day, but an additional 3–6 million requests originate from undeclared crawlers that impersonate a Chrome browser, rotate IP addresses and switch autonomous system numbers to bypass robots.txt files and Cloudflare firewall rules. Engineers created brand-new, non-indexed test domains that explicitly disallowed automated access and still found Perplexity could return detailed information from those sites. Cloudflare said the behaviour violates long-standing internet norms that require crawlers to identify themselves and honour site-owner preferences; compliant bots such as OpenAI’s were able to detect the blocks and ceased crawling, the company added. Perplexity disputed the findings, calling Cloudflare’s analysis a “sales pitch” and insisting the cited traffic does not belong to its service. The San Francisco-based start-up, founded in 2022 and valued at about $18 billion after a $100 million fund-raise last month, says its agents fetch content only in response to real-time user queries. Cloudflare has deployed new signature rules to detect the alleged stealth activity and is encouraging publishers to use its pay-per-crawl marketplace or free blocking tools to curb unwanted scraping.