A Federal Court Just Said AI Shopping Agents Can Browse Your Website. Here's What That Means for You

A federal appeals court just handed down a ruling that changes the rules for how AI agents interact with websites, and it happened this week.

Here's the short version. Amazon sued Perplexity last November over its Comet browser, an AI agent that can log into a user's Amazon account and complete purchases on their behalf. Amazon argued that counted as unauthorized computer access under a federal anti-hacking law. A district court agreed and blocked Comet from Amazon's logged-in pages back in March.

The Ninth Circuit just reversed that. The court's reasoning is the interesting part. It found that when an AI agent shops on a site because a user told it to, the user is the one accessing the site, not the software. The agent is just a tool carrying out instructions, similar to a browser or a mouse click.

That distinction matters way more than it sounds like it should.

Why this is a big deal

This is one of the first real court rulings anywhere that tries to answer a question nobody had to answer before: when an AI agent visits your website on someone else's behalf, who is actually the visitor?

Up until now, most businesses have treated AI agents like bots, something to block, throttle, or ignore. This ruling suggests that legal framing does not hold up cleanly when the agent is acting on direct instructions from a real customer. If a person tells an AI assistant to go check prices, compare reviews, or complete a purchase, the court is saying that is the person shopping, just with better tools.

The case is not fully over. It goes back to the lower court for more proceedings, and the appeals court was clear that its decision is based on the current record, not a sweeping new rule for all agentic AI everywhere. But the direction is unmistakable. Courts are starting to treat AI shopping agents as extensions of the customer, not as intruders.

What it means for a growing business

If you run any kind of online storefront, booking system, or customer portal, this is worth paying attention to now, not later.

• AI agents are going to show up in your traffic whether you invite them or not. Tools like this are built to shop, compare, and buy on a customer's behalf. Blocking them outright may not be the legal shield it used to look like.
• Being findable and readable by AI matters more every month. If an agent cannot easily understand your product pages, pricing, or checkout flow, it will send the customer somewhere it can. That is real revenue walking out the door.
• This is a preview of a much bigger shift. Agentic commerce, AI systems that browse, compare, and buy for people, is moving from experiment to normal. Retailers who treat this as a nuisance now will be scrambling to catch up in a year.
• Your terms of service and site design should assume an AI reader. That means clean structured data, clear pricing, and checkout flows that do not depend on a human squinting at a page.

This is not just an Amazon and Perplexity story. It is an early signal of how the legal system is going to treat AI agents acting on behalf of your customers, and it leans toward letting them operate.

The takeaway

AI agents shopping and transacting on behalf of real customers just got a legal green light, at least for now. Businesses that get ahead of this by making their sites and systems agent friendly will capture that traffic. Businesses that try to wall it off may be fighting a losing legal and practical battle. Either way, the smart move right now is to start treating AI agents as a real customer channel, not a bug to squash.