Major Websites Are Blocking AI Agents Acting on Users’ Behalf

Artificial intelligence (AI) agents developed by companies such as OpenAI, Meta, and Google are becoming increasingly popular. These systems perform tasks on users’ behalf, such as booking restaurant tables, cleaning up email, and purchasing products. However, major websites have begun restricting their access. In recent weeks, users of Muse, Grok, Dots, and Instinct have said they encountered blocks on popular services.

Yelp provides access only to AI agents that pay for its data licensing program. Delta Air Lines does not allow agents to book tickets on users’ behalf. JPMorgan Chase and Charles Schwab have confirmed that they block access to accounts because of security concerns. According to users, agents can also no longer access Cars.com and Amtrak, even though they previously worked with those services.

Last year, Amazon banned Perplexity’s agent-powered shopping feature and took the company to court. Amazon says third parties that purchase products on customers’ behalf must do so transparently and respect the site’s decision. A court initially ordered Perplexity to stop its agents from purchasing products on the site. However, an appeals court overturned that ruling in August. Amazon subsequently filed an amended complaint.

The dispute has not prevented other companies from launching their own agents. Nevertheless, Muse, Grok, Dots, and Instinct have also encountered similar problems on Amazon’s website. Amazon says Muse’s continued access to the site violates its rules. Instead of fighting the blocks, Meta has begun partnering with Walmart, Gap, Sephora, and Best Buy. Shopify has also opened its network of stores to AI agents.

Deliberate bans are not the only problem. Automated systems designed to prevent fraud and cyberattacks may identify AI agents as ordinary bots. Some users say their Google accounts were blocked for suspicious activity while they were using Muse. Others were unable to complete payments through Walmart despite the partnership. Such incidents make it difficult for agents to operate reliably on real websites.

In 2025, Cloudflare launched a feature that allows website owners to block AI crawlers that automatically read web pages with 1 click. The company later expanded its measures against systems that collect data for both search engines and AI training. However, as these blocks become more widespread, users may have to verify more frequently that they are not robots. This approach by websites also undermines the promise that AI agents will be able to perform any task on the internet.