Amazon and other retailers block AI agents from web access

Amazon is blocking rival shopping agents from accessing its website to protect customer data and adhere to access rules ^[1]. While personal AI agents like Meta's Muse are designed to navigate websites, fill shopping carts, and book travel, many platforms are implementing measures to stop them ^[1].

Key facts * Amazon is deliberately blocking certain tools from shopping on its site, while other websites use ordinary CAPTCHAs to stop AI agents ^[1]. * JPMorgan Chase blocks screen scraping and shared login credentials because they are described as "not secure" methods for transferring funds ^[1]. * eBay's user agreement prohibits unauthorized buy-for-me agents unless the company grants explicit permission ^[1]. * Walmart's experience shows that human-verification prompts can interrupt automated shopping sessions even when the checks were not intended to stop the assistant ^[1].

Why do websites block AI agents? Security is a primary reason for blocking AI agents, particularly when financial transactions are involved ^[1]. JPMorgan Chase's CFO Jeremy Barnum stated that some agents use shared login credentials and screen scraping to transfer funds, which are considered insecure methods ^[1]. Additionally, Amazon argues that certain agents violate access rules and put customer data at risk, a claim currently in a dispute with Perplexity over its Comet AI browser ^[1].

What commercial and technical factors influence access? Businesses often have commercial reasons to control automated access, such as eBay's prohibition on unauthorized buy-for-me agents ^[1]. Some restrictions are deliberate, while others result from security systems that cannot reliably distinguish legitimate agents from unwanted automated traffic ^[1]. For example, Zillow has acknowledged that its anti-scraping defenses sometimes catch agents it otherwise permits ^[1].

How do agents navigate these restrictions? AI agents can encounter human-verification prompts simply by navigating a website, even when following specific instructions from a user ^[1]. These checks can interrupt automated sessions, leading to a situation where the agent cannot finish the job and the user must return to the keyboard ^[1]. Switching on a VPN does not override these site-specific restrictions on automated access ^[1].

What are the standards for secure AI agents? NIST is working on the AI Agent Standards Initiative to ensure that agents capable of autonomous actions are adopted with confidence ^[2]. The initiative aims to foster industry-led technical standards and open protocols so that agents can function securely on behalf of users and interoperate across the digital landscape ^[2]. NIST also conducts research into agent authentication and identity infrastructure to enable secure human-agent and multi-agent interactions ^[2].