Cloudflare debuts "Disallow AI Training"

Cloudflare's move responds to a bind that publishers had been stuck in: AI crawlers scrape their content for training, but blocking those crawlers risks losing visibility in traditional search. By separating the two — training access from search indexing — Cloudflare is trying to make that trade-off unnecessary.

The compliance criteria matter as much as the tool. Apple, Google, and Microsoft have achieved or committed to the Accountable framework, giving the designation weight beyond Cloudflare's own network. Non-compliant mixed-use crawlers face network-level blocks, but only when a site owner chooses to enable the training block.

Commercial momentum is part of the story: Q2 2026 revenue of $696.1 million (up 36% year-over-year), a 120% net retention rate, and $4.16 billion in cash, with Q3 2026 revenue guidance of $736 million to $737 million. The unresolved tension is cost — GAAP gross margin slipped to 71.8% from 74.9% as network scale expanded. Whether the policy edge becomes durable pricing power likely hinges on how widely crawlers comply and whether margin pressure from network overhead eases.