Multiple Websites Restrict AI Agent Access

Recently, the boom in autonomous AI agents has run into website blocks. Companies such as OpenAI, Meta, xAI, Instinct, and Google have launched AI Agents that can access the internet on behalf of users to complete tasks like ordering food, shopping, managing email, and booking tickets, but multiple platforms have recently started restricting or blocking their access. Yelp stated that it only allows AI agents participating in data licensing programs to access its site; Delta, JPMorgan Chase, and Charles Schwab have prohibited agents from booking tickets or accessing accounts due to security, compliance, or user identity verification issues; websites previously accessible, such as Cars.com and Amtrak, have also become inaccessible. The controversy centers on the view that unauthorized AI agents may disrupt business rules, reduce ad exposure, and increase abuse risks, while companies like OpenAI and Meta emphasize that agents operate on behalf of users with their authorization, arguing that denying access is equivalent to denying the users themselves. The dispute between Amazon and Perplexity has highlighted this contradiction. Amazon claimed that third-party agent shopping violates its policies and repeatedly demanded that Perplexity stop related features; a court initially ordered Perplexity to cease these operations, but an appeals court overturned the order in August, after which Amazon filed an amended complaint. This year, Meta’s Muse was also cited by Amazon for violating site policies, and xAI’s Grok and OpenAI’s Dots were similarly reported as difficult to use on Amazon. Meta stated it prefers to collaborate with merchants to address concerns and has partnered with retailers such as Walmart, Gap, Sephora, and Best Buy, while Shopify has also opened its merchant network to AI agents. However, Google, Cloudflare, and anti-bot systems may still intercept agents due to "suspicious activity" or general protective mechanisms. Experts believe that while blocking can protect website security, it undermines the practicality of agents and forces users to face more human-verification challenges, making the internet experience more complex.