US appeals court rules AI agents act as the user, vacating Amazon's block on Perplexity

The US Ninth Circuit Court of Appeals on 4 August vacated the injunction that had blocked Perplexity’s Comet shopping agent from accessing Amazon since March. The court held that under the Computer Fraud and Abuse Act, it is the user – the person instructing the agent – who “accesses” a website, rather than the AI company operating the agent. The case returns to the lower court, and Amazon’s trademark and state-law claims continue, but the core anti-hacking theory for excluding AI agents has been rejected at appellate level for the first time.

Why this matters

For publishers and site owners deciding how to treat AI traffic, this ruling makes agent visits – the “ChatGPT-User” and “Perplexity-User” style fetches made on behalf of a real person – harder to litigate away, at least in the US. Practically, it strengthens the case for distinguishing between crawler types in your robots and firewall settings: training crawlers, search crawlers and user-directed agents are different things with different value exchanges. A site that blocks everything AI-shaped is now blocking traffic that courts are starting to treat as user traffic. For remote workers, it also protects the growing practice of having an AI assistant research visas, jobs or housing on your behalf.

Context

The ruling cuts across a litigation wave otherwise running against Perplexity – Reddit’s copyright suit survived dismissal in late July and CNN filed its own claim days earlier – but those cases concern content copying for training and answers, a separate question from whether an agent may fetch a page for a user at all. In Europe, Germany’s media regulator went the other direction in July, bringing AI search answers under editorial-content rules. For sites behind Cloudflare, note that new AI-crawler defaults take effect on 15 September; anyone relying on AI visibility should check that search and agent categories remain open.