How GitBook serves 30,000 sites with sub-second content updates
41% of traffic isn't human. GitBook's caching architecture now has to handle AI crawlers sweeping every page across 30,000 sites—and still guarantee sub-300ms updates.
Why it matters
As AI crawlers become the majority traffic source for developer documentation platforms, infrastructure design must shift from human-centric patterns to handle unpredictable, high-volume programmatic access while maintaining consistency and cost predictability at scale.
The key facts
8 to know30,000 documentation sites hosted on Vercel
120 million monthly page views
40,000 cache invalidations processed daily, each resolved in under 300ms
41% of all traffic now comes from AI crawlers and automated systems
AI-driven traffic surged 5x year-over-year in 2025
Tag-based cache invalidation triggered by merge events across GitHub, GitLab, and GitBook
Updated content becomes visible globally within 300 milliseconds after merge
Infrastructure designed for ~100% cache hit targets to maintain latency and predictability
Go to the source
Vercel Blogvercel.com
Publisher excerpt: GitBook on Vercel 30,000 documentation sites hosted on a single Vercel deployment 120 million monthly page views served from the edge 40,000 cache invalidations processed daily, each resolved in under 300ms 41% of all traffic now comes from AI crawlers and automated systems GitBook hosts 30,000…