Websites are increasingly blocking AI agents from completing tasks

AI agents from major technology companies are encountering growing resistance from the websites they are meant to use on behalf of people. Travel, finance, retail, and review platforms are restricting automated access, limiting promises that agents can book tickets, shop, or manage accounts across the web.

Emma Roth reports for The Verge that Yelp permits AI agents only through its paid data licensing program, while Delta Air Lines prevents agents from booking flights. JPMorgan Chase and Charles Schwab have also confirmed blocks on agent access to customer accounts, citing security concerns.

Amazon remains a central battleground. The company previously banned Perplexity’s shopping agent and sued the startup. Amazon argues that third parties buying products for customers must identify themselves and respect a site’s decision not to participate. Perplexity counters that an agent acts as a user’s assistant. A court order initially stopped the service, but an appeals court later reversed that ruling. Amazon has filed an amended complaint.

Security systems create another barrier

Formal bans are only part of the problem. AI agents can also trigger systems designed to stop fraud, scraping, and other automated abuse. Users have reported account blocks and failed checkouts when using agents, including on services that have commercial partnerships with AI companies.

Meta has responded by working with retailers including Walmart, Gap, Sephora, and Best Buy for its Muse assistant. Shopify has also opened its merchant network to AI agents. These agreements may create more reliable routes for agents, but they also point toward a web where access depends increasingly on business deals.

Cloudflare’s anti-bot tools add further pressure. The company has introduced options for websites to block AI crawlers and has targeted bots that combine search indexing with AI data collection. Chris Canal, chief executive of AI safety evaluator EquiStamp, tells The Verge that wider restrictions could reduce the practical value of agents and lead to more verification steps for human users.