While the retail world fixates on which AI assistant gets named when a shopper asks where to buy, a courtroom in Seattle is deciding something more fundamental: whether those assistants are legally allowed through the door at all. On Thursday, the Ninth Circuit Court of Appeals hears oral arguments in Amazon.com Services LLC v. Perplexity AI, Inc.the first federal appellate test of whether an AI agent acting on a user's explicit instruction counts as an authorized visitor to a logged-in commercial website.

What the fight is actually about

The case centers on Perplexity's Comet browser, which can log into a shopper's own Amazon account and complete a purchase on their behalf. Amazon argues that conduct violates the federal Computer Fraud and Abuse Act and California Penal Code §502 — even when the user explicitly told the agent to do it. In March, U.S. District Judge Maxine Chesney granted Amazon a preliminary injunction, finding the company likely to prevail, before the court issued an administrative stay weeks later. Now it goes up on appeal.

Strip away the statutes and the question is almost philosophical: if I authorize software to act as me, is that software me — or is it an unauthorized third party wearing my credentials? Amazon says the latter. Perplexity, and a notable lineup of civil-liberties groups, say the former.

Why retailers can't sit this one out

It's tempting to read this as an Amazon-versus-startup turf war, and there's plenty of self-interest on both sides. Amazon has its own agents — Rufus, Alexa+, "Buy for Me" — and a clear incentive to keep rival bots from transacting inside its walls while its own roam free. But the precedent won't stay contained to Amazon's storefront.

If the Ninth Circuit affirms that a retailer can invoke the CFAA to block an agent operating under a customer's direct instruction, then every retailer with a logged-in experience gets a powerful tool to decide which AI intermediaries may transact on their site — and on what terms. If the court goes the other way, the logged-in storefront effectively becomes open territory for any agent a customer points at it, and the leverage shifts to whoever controls the assistant. As the legal scholar Eric Goldman put it in a closely-read analysis of the case, the outcome will shape the ground rules for agentic commerce far beyond the two named parties.

The legal threads to watch

Court-watchers expect the panel to press three points: how the Supreme Court's Van Buren decision applies to agent traffic operating under user authorization; whether the precedent from Facebook v. Power Ventures — built for third parties scraping a platform — stretches to an agent acting on a logged-in user's direct command; and whether the §502 claim survives a First Amendment challenge raised by the ACLU and the Knight First Amendment Institute.

That last thread is the sleeper. If a court accepts that automating your own account access implicates expression, it complicates every "no bots" term of service in retail.

The takeaway

This is the part of the agentic-commerce buildout that doesn't show up in a glossy keynote. While platforms race to publish open shopping protocols and stuff checkout into chat windows, the question of who is even allowed to press buy remains legally unsettled — and a three-judge panel is about to give the industry its first real answer. Retailers betting their roadmaps on a frictionless agent future should be reading the docket, not just the press releases.

A ruling is not expected immediately; appellate decisions typically follow oral arguments by weeks or months. We'll cover the opinion when it lands.