个人AI代理在购物、订票等场景遇网站主动拦截与反机器人机制，用户被卡在中间。
Web-based AI agents are increasingly blocked by purpose-built anti-bot systems: dynamic CAPTCHAs, behavioral fingerprinting, IP rate limiting, and JS challenges—not accidental failures, but intentional commercial safeguards. As TechCrunch reports, these blocks leave end users stranded mid-task, revealing a hard infrastructure constraint: automation access is no longer technical, but negotiated. For early AI application teams, this invalidates 'scrape-first' development. Sustainable paths require higher upfront cost: integrating official APIs (e.g., Amadeus, Sabre), co-building whitelisted channels with SaaS partners, or investing in robust browser automation (real-device fingerprinting, distributed proxy rotation, human-in-the-loop fallback). Yes, it increases engineering overhead—but avoids legal exposure, unpredictable downtime, and builds trust with platform providers instead of fighting them.