Building unique, per-customer defenses against advanced bot threats in the AI era

Cloudflare is unveiling a per-customer approach to bot defence that uses behavioural anomaly detection unique to each bot management customer, delivering hyper-personalized protections as AI-powered scraping grows more sophisticated.

The new platform deploys a three-step process to learn what normal looks like for each website or API: establishing a dynamic, zone-specific baseline, identifying deviations from that baseline, and generating actionable findings that can influence Bot Scores and trigger targeted responses. This per-customer model supports millions of detections by leveraging hundreds of signals, including HTTP/2 fingerprints and Client Hello extensions, with over 50 heuristics developed since June to catch evolving bots.

In practice, early results show hundreds of millions of scraping requests detected in a 24-hour beta window, with 138 million scraping requests flagged across just five zones, and 34% of these detections not would have been caught by the existing bot-score system. Cloudflare notes the global importance of the approach as AI-driven scraping now accounts for nearly 80% of AI bot activity in mid-2025, highlighting the need for bespoke, contextual protections that adapt to each site’s normal traffic.