There's a federal lawsuit underway that sounds like a tech trade dispute but is actually about something much more fundamental: who decides which AI tools your customers are allowed to use when they shop on Amazon. The answer, at least for now, is Amazon. And the legal theory backing that answer has real implications for where your brand needs to show up over the next few years.
The case is Amazon v. Perplexity, and it's worth understanding even if you've never used an AI shopping agent in your life.
What Actually Happened
Perplexity built a browser called Comet that lets users ask an AI assistant to find and buy things on their behalf, including on Amazon. Users log into their own Amazon accounts, give Comet permission to act for them, and Comet does the shopping. From a consumer standpoint, this is convenient. From Amazon's standpoint, it's unauthorized access to a protected computer system.
Amazon sued Perplexity in November 2025 under the Computer Fraud and Abuse Act, a 1986 anti-hacking statute originally written for a world that had never heard of an AI agent. Amazon's argument: it never authorized Comet to access its platform, and a user's permission doesn't substitute for platform permission. In March 2026, a federal judge agreed and issued a preliminary injunction blocking Comet from Amazon's logged-in pages.
The Ninth Circuit stayed that injunction pending appeal, and oral arguments happened on June 11. No ruling yet. But the district court's logic is already on the record, and it's the part that matters for brand sellers.
User Permission Is Not Platform Permission
Judge Chesney's ruling drew a line that no court had drawn quite so clearly before in the AI context: a user authorizing an AI agent to act on their behalf does not mean the platform has authorized that agent. She called it "access with the Amazon user's permission, but without authorization by Amazon." Those two things are legally distinct.
The analogy that keeps showing up in the coverage is a good one. Giving a friend your apartment key doesn't give them access to the building's common areas. Your authorization covers your door, not the building's systems. By the same logic, your customers can tell an AI agent to shop for them all they want, but that doesn't mean Amazon has agreed to let the agent in.
Perplexity's counterargument is that the agent is just the user's browser, and the user is the one accessing Amazon, not Perplexity. It's a reasonable position, and the Ninth Circuit's stay suggests the appellate panel isn't certain Amazon is right. But the district court's ruling stands as the current precedent, and it leans Amazon's way.
The Double Standard That's Hard to Ignore
Here's where it gets interesting. Amazon's Buy for Me feature does, functionally, the same thing in reverse. Amazon's AI agent browses independent retailers' websites, scrapes product data, lists those products on Amazon, and completes purchases on customers' behalf.
Retailers were included by default and had to opt out by emailing Amazon directly. Over 500,000 products from independent sites were pulled in without consent. So Amazon is using federal hacking law to block other companies' agents from its platform, while simultaneously sending its own agents onto other companies' platforms. Completely unbiased, as you'd expect.
Courts may eventually sort this out. For now, the practical reality is that Amazon controls who gets to agent on Amazon, and that's not changing soon.
What This Means for Your Discoverability Strategy
The CFAA case isn't really about Perplexity. It's about the architecture of agentic commerce: which platforms build open, authorized access channels for AI agents, and which ones play enforcement games. Those two strategies produce very different discovery surfaces.
Amazon's approach is clear. It controls the agents on its platform. Alexa for Shopping, Buy for Me, Sponsored Prompts, the agentic ad formats we've covered in recent posts , all of these are Amazon's AI, operating inside Amazon's walled garden, surfacing products on Amazon's terms. If you want to be found by Amazon's agents, you play by Amazon's rules: optimize your listings, run your ads, and make sure your catalog data is clean.
Outside Amazon, the picture is different. Platforms that build open agent access channels, through APIs or structured data, are where third-party AI agents can actually operate. If a shopper's AI assistant can buy from one retailer but gets legally blocked from another, the open one wins the agent-mediated sale every time.
Three Things to Do Right Now
First, treat your Amazon catalog data as the foundation of agent-mediated discovery. Agents on Amazon's platform surface products based on structured signals: titles, bullet points, A+ content, reviews, and ad bids. If your listing data is thin or inconsistent, you'll perform badly whether a human or an AI is doing the shopping. The same content optimization principles that help with AI search visibility apply here.
Second, pay attention to which platforms are building authorized agent access versus which ones are walling themselves off. Amazon's garden is walled, but it's also enormous, and Amazon is building a lot of AI infrastructure inside it. Platforms that build open protocols are where agent-mediated discovery will grow outside of Amazon. Both matter, for different reasons, and your investment should reflect that.
Third, don't wait for the Ninth Circuit to tell you how this ends. The ruling will set precedent for the western United States. But the underlying dynamic, platforms controlling which AI tools can access them, is already operating regardless of how the case turns out. Build your discoverability strategy around that reality.
The Bigger Picture
The Amazon v. Perplexity case is the first major legal test of a question that's going to define e-commerce for years: is an AI agent shopping on your behalf an extension of you, or a third party that needs separate platform authorization? The district court said third party. The Ninth Circuit hasn't decided yet.
What's clear is that the platforms building the most sophisticated AI shopping infrastructure, and building the legal moats around it, are going to have outsized control over where products get discovered. Amazon is both of those things at once. That's not a reason to panic. It's a reason to be very deliberate about how you're positioned inside its ecosystem.
If you'd like to talk through what your catalog looks like to Amazon's agents right now, and where there are gaps to close before the agentic commerce landscape shifts further, we'd be glad to take a look. Schedule a call and let's dig in.