
Reverse Engineering "Verified Bot": The Six-Layer Identity Chain That Decides Whether Googlebot Sees Your Site (and the September 15, 2026 Default That Quietly Rewired It)
On September 15, 2026, Cloudflare changed what a checkbox means. Sites that had told it to block AI training started blocking Googlebot, Applebot, and Bingbot on their ad pages — not because anyone chose to block Google, but because those crawlers are classified as Search + Training and the most restrictive rule wins. This is the reverse-engineering teardown underneath that headline: the six layers that actually decide whether a request is "Googlebot," what each layer proves versus merely claims, the seven ways the new signed-bot standard fails in production (including an SSRF hole nobody warns you about), a reproducible six-command protocol to test your own site, the eight-state access matrix, and the classifier code to replace user-agent matching with evidence.











