The blocks are not always intentional. Anti-scraping defences and CAPTCHA systems cannot always tell a legitimate agent from unwanted automated traffic, and Zillow has acknowledged its defences sometimes catch agents it otherwise permits. Google notes that shared VPN networks can trigger reCAPTCHA challenges for similar reasons, so an agent can trip security checks just by navigating a site.

Money raises the stakes. On JPMorgan Chase's October earnings call, CFO Jeremy Barnum said some agents use shared login credentials and screen scraping to move funds, methods he described as "not secure," and warned that customers who hand over their credentials could be responsible if an agent takes an unwanted action. The bank blocks those methods where possible.

Commercial interests are in play too. Amazon is fighting Perplexity's Comet browser in court, arguing it violates access rules and risks customer data, while Perplexity says users have authorised it; a federal appeals court vacated a preliminary injunction against Perplexity in August 2026, though the dispute continues. eBay's user agreement prohibits unauthorised buy-for-me agents, and Delta told TechCrunch it has no integration for third-party agents to book flights. Notably, Amazon runs its own Buy for Me feature that buys from other brands' sites.

Infrastructure is adapting as well. Cloudflare now lets site owners treat search crawlers, AI training bots and user-directed agents differently, and since September 15 new domains on its network default to allowing search crawlers while blocking AI training bots and agents on pages showing ads. Shopify, meanwhile, allows compatible browser agents to update checkout details and submit orders after buyer confirmation.