Court Rules AI Shopping Agents Are Legal in Amazon v. Perplexity

On August 4, 2026, a federal appeals court issued the first ruling of its kind on a question every major retailer is now watching: can an AI agent lawfully act inside a user's private account on their behalf? On the current record, the Ninth Circuit said yes. Amazon had $19.8 billion in quarterly advertising revenue tied to the answer going the other way.

The court vacated a preliminary injunction that had blocked Perplexity's AI shopping tools from accessing Amazon accounts. The ruling did not establish a general legal right for AI agents to shop on any platform. It held, on the specific facts before the panel, that Amazon was unlikely to prove that Perplexity itself had accessed Amazon's computers under federal computer-access law. The underlying lawsuit continues. But the decision sends a clear early signal about where the courts are leaning, and it arrives while every major payments company and retailer is already building the infrastructure to support this kind of agent-driven commerce.

Who Is Perplexity and What Is Comet?

Perplexity is an AI-powered answer engine founded in 2022. Rather than returning a list of links, it reads the web in real time and delivers a single cited answer. As of April 2026, it had more than 100 million monthly active users and a reported $20 billion valuation at its last funding round, according to the Financial Times. Among its investors is Amazon founder Jeff Bezos personally, which adds a notable layer to this lawsuit.

Comet is Perplexity's AI-powered browser. It navigates websites and can take action on a user's behalf, including logging into a user's Amazon account using their own stored credentials and completing a purchase without the user touching the keyboard. The user sets the parameters. The agent executes. That is what Amazon sued over, and that is what the Ninth Circuit's ruling addresses.

How This Case Unfolded

Late 2024: A Pause, Then a Dispute

According to Amazon, the two companies had previously agreed to pause agentic shopping on Amazon's platform in November 2024. Amazon alleged that Perplexity re-enabled the feature anyway and that it failed to identify Comet's automated activity as an AI agent, making it difficult to distinguish from ordinary browser traffic. Perplexity disputed aspects of that characterization.

November 2025: Amazon Files Suit

Amazon filed suit under the Computer Fraud and Abuse Act and California's comparable computer-access statute, alleging Perplexity accessed its systems without authorization, created privacy vulnerabilities, and interfered with its platform. Perplexity characterized the suit as an attempt to limit consumer choice in AI tools.

March 9, 2026: District Court Grants Amazon an Injunction

Senior US District Judge Maxine Chesney sided with Amazon and issued a preliminary injunction. The order also required Perplexity to destroy any Amazon data it had collected through the agent. Perplexity appealed the following day. The Ninth Circuit heard oral arguments in Seattle on June 11 and issued its ruling on August 4.

August 4, 2026: The Ninth Circuit Reverses the Injunction

The panel vacated the injunction, finding that Amazon was unlikely, on the record before the court, to prove that Perplexity itself had accessed its computers. The court found it was the Amazon user, using Perplexity's agent as a tool, who accessed the platform. The court expressly noted its holding is limited to the CFAA and CDAFA access question as applied to this agent and this platform on the existing record. It also specifically preserved Amazon's ability to regulate access through its private terms of service. The underlying lawsuit returns to district court.

Why Amazon Is Fighting This So Hard

Amazon publicly framed its case around access, security, and platform integrity. Its advertising business gives it a significant additional commercial stake in how third-party agents interact with its marketplace.

The Alexa for Shopping Contrast

Amazon's position is complicated by what it has been building inside its own platform. In May 2026, Amazon brought its Rufus shopping assistant and its Alexa+ assistant together into a unified product called Alexa for Shopping, reported by GeekWire and Axios. The assistant can now track prices and complete purchases automatically when items hit a target price, set up recurring purchases, and add items to a cart based on user preferences. Amazon reported that more than 350 million customers used the assistant in the preceding 12 months. Amazon also reported in July 2026 that US customers who used Alexa for Shopping spent more than 40 percent more per order on average than those who did not, though that reflects an association rather than proven causation.

The distinction Amazon would draw is meaningful: its agent operates within its own ecosystem and keeps purchases on its platform. Perplexity's agent can act across the open web using a user's own credentials. Amazon is building many of the same agentic purchasing capabilities inside its ecosystem while challenging the way an external company's agent accesses its platform.

How AI Agents Are Already Shopping for Your Products

According to the IAB, nearly 40 percent of US shoppers already use AI while shopping, and 80 percent of that group expects to use it more. McKinsey projects that US consumer retail could see up to $1 trillion in revenue orchestrated through agentic commerce by 2030. Gartner forecasts that by 2028, 90 percent of B2B buying will be intermediated by AI agents, moving more than $15 trillion through agent exchanges.

What This Means for Your Product Visibility

AI shopping systems can discover and evaluate products through several routes, including rendered webpages, screenshots, search indexes, merchant feeds, APIs, and structured metadata, depending on the product and platform. Comet itself works in part by capturing screenshots of what it encounters in the browser. Clear, complete, machine-readable product information can make comparison easier for these systems. Schema markup is one useful form of structured metadata, but it is one part of a broader picture that includes accurate titles, specific use-case descriptions, consistent pricing, and clean availability data.

For merchants selling on Amazon, Alexa for Shopping is already changing how products are surfaced inside the platform. According to Tinuiti's analysis of the assistant's recommendation logic, the system interprets intent and filters products through signals that go beyond keyword matching. Complete, accurate listing content is now a discovery asset alongside paid placement.

The Legal Road That Built This Ruling

Two prior cases shaped the framework the Ninth Circuit applied here. In Van Buren v. United States (2021), the Supreme Court narrowed the CFAA by establishing that the statute applies to whether someone may enter a computer system or particular areas within it, not to cases where someone misuses access they are otherwise entitled to have. Separately, in hiQ Labs v. LinkedIn (2019 and 2022), the Ninth Circuit twice held that automated scraping of publicly available data did not violate the CFAA. As analyzed by both Fenwick and Proskauer, the reasoning was that federal computer-fraud law was not designed to give platforms broad control over who accesses their systems.

What made this case different from hiQ was the password-protected element. Comet accessed private user accounts, not public pages. That is why the district court initially sided with Amazon. The Ninth Circuit applied the Van Buren framework to that more sensitive context and still found that user authorization was the legally relevant factor on the record before it. Platform approval, as a separate requirement, was not established by the law. The Electronic Frontier Foundation, which filed a brief supporting Perplexity, called the outcome correct, noting that "that's the right conclusion, as both a legal and technical matter."

Amazon still has options. The court's opinion explicitly preserves Amazon's ability to enforce its private terms of service, which is a separate legal theory this ruling does not resolve. Amazon could also seek en banc rehearing from the full Ninth Circuit or petition the Supreme Court, though neither path is fast or guaranteed.

What This Means for Your Business Data and Transactions

When a human customer shops your store, they generate behavioral data at every step. You can see what they clicked, what they abandoned, and how long they stayed on each page. Agentic purchasing may complicate attribution because some or all of the research and decision process can happen outside a merchant's normal site journey. Depending on the platform and integration, merchants may receive different or more limited referral and session signals.

The Emerging Question Around Disputes

Agentic transactions also introduce emerging questions about how merchants prove user authorization when an AI agent acts on a customer's behalf. Existing payment and chargeback frameworks still apply. However, their application to delegated AI transactions is developing. According to research from Chargeflow on agentic commerce regulation, Regulation E, the federal law governing electronic payment disputes, defines authorization as the consumer granting permission for a transfer. Whether a consumer's general instruction to an agent constitutes authorization for any specific purchase it makes is not yet settled.

A Darwinium survey of 500 fraud and risk professionals found no consensus on who should bear liability when an agent-driven transaction goes wrong. Thirty-nine percent said the AI provider, 20 percent said the consumer, and 14 percent said the merchant. Card networks have not updated their dispute rules to account for agent-initiated purchases. According to guidance from Chargeflow, agent authorization logs and delegation records are becoming the primary evidence layer for merchants defending disputes where the buyer was an AI agent.

How the Rest of the Industry Is Responding

Amazon's strategy is to restrict external agents while building its own. The broader industry is taking a different approach.

In January 2026 at the National Retail Federation conference, Google CEO Sundar Pichai and John Furner, then incoming President and CEO of Walmart Inc., announced a partnership bringing Walmart and Sam's Club shopping directly into Google's Gemini assistant. At the same event, Google introduced the Universal Commerce Protocol, an open interoperability standard for AI agents built with Walmart, Target, Shopify, Etsy, Wayfair, and more than 20 other companies. The protocol is designed to allow AI agents to operate across discovery, checkout, and post-purchase support. Providing interoperability does not guarantee unrestricted access by every agent, and specific implementations will vary by retailer. But the strategic direction it signals stands in sharp contrast to Amazon's litigation approach.

Perplexity itself is also worth understanding as a discovery channel independent of the lawsuit. With more than 100 million monthly active users, it functions as a search engine that gives direct cited answers rather than link lists. For small businesses, being discoverable on Perplexity is a meaningful third channel alongside Google and Amazon, and one that operates on the quality and clarity of your content rather than paid placement.

The Bottom Line for Small Business Owners

The Ninth Circuit's ruling is a narrow one. The court said so explicitly. It does not establish a broad legal right for AI agents to shop on any platform. But it is a meaningful early signal that courts are treating agent-driven activity as an extension of the user's own authorized access rather than a new category of unauthorized intrusion. That framing, even on a limited record, is likely to shape how similar disputes are decided going forward.

The practical implications for small businesses are not narrow at all. How customers discover your products is changing. How your transaction data is generated is changing. How you document authorization when an AI agent completes a purchase on a customer's behalf is becoming a meaningful operational practice. Businesses that adapt early will have cleaner data, fewer disputes, and better visibility in AI-driven search and shopping systems than those that wait for the legal landscape to fully settle.

That kind of adaptation requires capital. Improving product data infrastructure, updating checkout flows, building documentation systems around agent-driven transactions, none of it happens without runway. Fund&Grow helps entrepreneurs access up to $3000,000 in business credit cards with zero percent introductory interest rates, giving you the working capital to make these investments on your own timeline without drawing down reserves or taking on fixed debt service at the exact moment your business needs flexibility most.

Check Your Lending Potential

Copyright © 2026 Fund&Grow. All rights reserved. This article contains Fund&Grow commentary based on cited public and third-party sources. Underlying data remains attributable to the original sources cited.