A federal court conflict between Amazon and Perplexity AI over Perplexity’s Comet browser and its optional AI shopping assistant has produced a preliminary injunction and an expedited appeal that could affect how client-side AI agents interact with commercial websites and how the federal Computer Fraud and Abuse Act (CFAA) is applied.

Amazon sued Perplexity in U.S. federal court, alleging that Comet’s assistant accessed Amazon customer accounts and Amazon’s systems without authorization, impersonated a Google Chrome browser, masked its digital fingerprint to evade blocks, continued accessing the platform after warnings, and thereby harmed Amazon’s advertising business. Amazon told the court it spent more than $5,000 and numerous employee hours developing tools to block the browser and prevent future unauthorized access, and it said automated agents created security risks by acting within protected systems, including areas requiring passwords. Amazon invoked the CFAA and California’s analogous statute and sought injunctive relief to stop the assistant from accessing the site.

The district court granted a preliminary injunction barring Perplexity’s assistant from accessing Amazon accounts on March 9, 2026, while expressing judicial doubt about whether the CFAA cleanly fits the conduct at issue. The judge included a one-week stay to allow Perplexity to appeal. The Ninth Circuit stayed the injunction pending an expedited appeal and set an accelerated schedule.

Perplexity has appealed and asked the Ninth Circuit to lift the injunction. In its filings Perplexity says the Comet assistant is a browser-based tool that users activate to browse Amazon from their own devices, that it captures screenshots or HTML snapshots of a user’s session, encrypts and sends those snapshots to Perplexity’s servers for processing, then returns instructions to the browser without Perplexity’s servers directly connecting to Amazon, and that screenshots are deleted within 30 days. Perplexity argues those technical details mean Perplexity did not itself access Amazon’s computers; that users expressly authorized the assistant to access their account data; and that Amazon has produced no evidence of concrete harms such as lost sales, reduced traffic, or dissatisfied customers during roughly eight months Comet operated. Perplexity has described the lawsuit as an attempt to restrict users’ ability to choose AI tools and said it will defend the product in court.

Perplexity’s appeal brief argues the district court improperly conflated factors required for injunctive relief and that Amazon failed to show technological loss, intentional direct access by Perplexity, or lack of user authorization. Amazon contends the assistant bypassed sponsored product listings, recommendations, and other advertising placements that produce the majority of Amazon’s ad revenue and that blocking such agents is necessary to preserve advertising impressions and billing integrity.

Legal analysis of the dispute identifies two central doctrinal questions: whether the CFAA covers software that operates on a user’s device and only receives information already delivered to that user’s browser, and when actions by an AI acting on a person’s behalf create liability for the AI developer or the platform user. Commentators note the case implicates prior decisions shaping web-access law, including Ninth Circuit scraping litigation and the U.S. Supreme Court’s Van Buren v. United States, which narrowed the CFAA by distinguishing authorized access from bypassing explicit technical barriers. Observers also point to alternative legal theories Amazon could pursue, including state tort claims, contract-based remedies, misappropriation, unjust enrichment, conversion, breach of contract, and privacy causes of action.

The outcome has wider commercial implications. Amazon argues a ruling for Perplexity would undermine an advertising model that depends on human visibility of sponsored listings and search-based placements that helped generate about $68.6 billion in annual ad revenue for Amazon. Perplexity and other commentators say a ruling for Perplexity would limit the CFAA’s reach over locally run AI assistants and shift disputes toward technical countermeasures and contractual rules. Major platforms have taken differing approaches: some are developing open protocols for AI-mediated commerce, while Amazon has adopted contractual Agent Policies and has blocked multiple external agents as it rolls out its own shopping assistants, which it reports have reached hundreds of millions of users and produced significant incremental sales.

The Ninth Circuit’s expedited decision will determine whether a statute enacted in 1984 can be used to restrict client-side AI assistants that interact with commercial websites and will shape how platforms, developers, and users allocate control and responsibility for autonomous shopping agents.

Original Sources: 1, 2, 3, 4, 5, 6, 7, 8 (amazon) (comet) (chromium) (platforms) (stay)

Short answer: The article is mostly news and legal analysis; it offers little direct, actionable help for an ordinary reader. It explains the stakes and positions clearly enough to inform, but it does not provide practical steps, tools, or concrete guidance someone can use right away.

Actionable information The article does not give clear steps, choices, or instructions a normal person can act on today. It describes what each side claims, the judge’s posture, and the potential industry implications, but it does not tell readers how to protect their accounts, change settings, switch services, or otherwise respond. References to industry options—platform policies, protocols, technical countermeasures—are descriptive rather than prescriptive. There are no links, checklists, or tools a reader could follow or apply immediately. In short, there is nothing a typical reader can do tomorrow that the article explains how to do.

Educational depth The piece provides useful context about the legal argument (how Perplexity says its agent operates, and how Amazon frames the harm) and frames the dispute in terms of a broader industry question about control of AI agents. That gives some explanatory value beyond surface facts. However, it stays at a high level: it does not explain the Computer Fraud and Abuse Act’s legal standards in practical terms, it omits technical detail about how client-side assistants, screenshots, or encrypted forwarding actually work, and it does not analyze likely legal standards the Ninth Circuit might apply. Numbers are minimal (Amazon’s $68.6 billion ad business is cited) but the article does not break down how that figure connects to ad impressions or revenue mechanics. So it teaches the outline and stakes, but not the deeper legal tests, technical mechanics, or economic models in a way that lets a reader fully understand cause and effect.

Personal relevance For most readers this is background news: interesting if you follow tech law, e-commerce, or AI, but not directly affecting daily safety, health, or immediate financial decisions. It is more relevant to three groups: users of AI browsing assistants who care about privacy and account behavior, developers building such agents, and advertisers or e-commerce businesses that rely on page impressions. If you are none of those, the relevance is limited. The article does not make clear what an ordinary Amazon customer should change about their habits or settings, so it fails to translate the legal fight into personal actions.

Public service function The article does not provide warnings, emergency information, or public-safety guidance. It recounts a legal dispute and its industry implications but does not tell readers how to act responsibly, protect themselves, or evaluate services. As a result it functions primarily to inform rather than to advise or protect the public.

Practical advice There is effectively no practical advice an ordinary reader can follow. The piece does not recommend steps such as auditing browser extensions, disabling optional assistants, checking account activity, or reviewing privacy settings. Any suggestions about contractual approaches or technical countermeasures are discussed at an industry level and are not actionable for most people. If a reader wanted to respond personally, the article does not equip them.

Long-term impact The article points to potentially large long-term consequences: a court ruling could change how platforms control third-party AI agents and affect online advertising models. That gives readers foresight about industry direction, which can help professionals plan. But for most individuals it does not offer durable guidance to improve habits, prepare for changes, or avoid repeating mistakes. The value is mainly informational: know that this legal issue might reshape ecommerce, but no instructions follow.

Emotional and psychological impact The article is measured rather than sensational. It does not appear designed to provoke fear or shock; it lays out opposing claims with some legal skepticism mentioned. For readers worried about being surveilled by AI or about ad-driven platforms changing, it may raise concern without giving coping steps, which can produce mild unease. But it does not appear exploitative or sensationalist.

Clickbait or ad-driven language The article reads like straightforward reporting and legal analysis. It does not rely on dramatic exaggeration or overpromising. The framing of stakes and possible consequences is appropriate to the subject.

Missed chances to teach or guide The piece missed several opportunities to be more useful. It could have: • Explained, in simple terms, what the Computer Fraud and Abuse Act requires and how that might apply to client-side tools. • Described concrete technical ways an AI assistant can interact with a website and why those differences matter legally and operationally. • Given practical privacy and security steps users can take when using AI browsing assistants. • Outlined what developers, advertisers, or small businesses should watch for and prepare to do if platforms change enforcement. Because the article left those out, readers who want to act or learn more must look elsewhere.

Practical, real-world guidance you can use now If you are an ordinary user worried about AI-assisted browsing or about third-party agents acting on your behalf, take these simple, realistic steps. Check the permissions and settings of any browser or assistant before activating it, and disable optional assistants you do not trust. Review your browser extensions and uninstall ones you do not recognize; many assistants run as extensions or bundled features. Monitor your account activity logs on services you use regularly (Amazon and other major sites provide order and login histories) and enable multi-factor authentication to reduce risk from unauthorized automation. When using an assistant that captures pages or screenshots, assume anything visible in your browser could be transmitted and avoid performing sensitive transactions while it is active. For developers or small businesses that rely on advertising impressions, prepare basic contingency plans: document how your ad billing works, track impression metrics so you can spot drops quickly, and include clauses in contracts that define acceptable bot and agent behavior. If you are an advertiser purchasing placements, ask your platform for clarity on agent policies and request reporting that distinguishes human views from agent-driven interactions. For anyone trying to evaluate reporting about this topic, compare multiple reputable outlets, look for direct court filings or reputable legal analysis to confirm claims, and be skeptical of articles that make sweeping claims about "end of advertising" consequences without explaining mechanisms. These steps do not require technical expertise and help you reduce immediate risk, preserve account security, and stay informed as legal developments unfold.

"Amazon contends that the assistant lets Perplexity bypass sponsored product listings, recommendations, and other advertising placements that generate the majority of Amazon’s ad revenue, and that blocking such agents is necessary to preserve advertising impressions and billing integrity." This sentence frames Amazon’s claim in a way that privileges Amazon’s commercial harms as the central problem. It highlights "majority of Amazon’s ad revenue" and "necessary to preserve," which push readers to see blocking agents as justified. That choice of strong commercial-impact words helps Amazon’s position and downplays Perplexity’s counterarguments.

"Perplexity responds that users authorized the assistant to access their own account data, that no private information beyond what a user sees in a browser was transmitted, and that the complaint concerns lost advertising impressions rather than the kind of technological harm the statute addresses." Calling the harm merely "lost advertising impressions" minimizes Amazon’s framing and uses a softer phrase to suggest non-seriousness. The contrast set up by "rather than the kind of technological harm the statute addresses" guides readers to treat Amazon’s legal theory as a mismatch, favoring Perplexity’s legal position.

"The district court issued a preliminary injunction despite expressing doubt that the statute was a good fit for the conduct at issue, and the Ninth Circuit stayed that injunction pending expedited appeal." Saying the district court "expressing doubt" about the statute's fit inserts the judge’s skepticism into the narrative and signals uncertainty about Amazon’s legal claim. This placement can bias readers to view Amazon’s theory as weak even though an injunction was issued, subtly undermining the injunction’s weight.

"Perplexity’s appeal brief argues the district court improperly conflated the distinct requirements for injunctive relief and that Amazon failed to show technological loss, intentional access by Perplexity itself, or lack of user authorization." Listing Perplexity’s contentions in plain terms without Amazon’s rebuttal gives readers a one-sided sense of the appeal. The absence of counter-phrasing or evidence supporting Amazon’s showing makes Perplexity’s legal arguments appear more persuasive by omission.

"The dispute highlights an industry-wide question about who controls AI agents that act on users’ behalf when visiting commercial websites and whether a decades-old anti-hacking law applies to client-side AI tools." Describing the law as "decades-old" and "anti-hacking" frames it as potentially outdated for modern AI, which nudges readers to view its application skeptically. That wording supports the idea that existing laws may not fit new technology, aligning the narrative with Perplexity’s position.

"Major platforms have taken differing approaches: some are building open protocols to enable AI-mediated commerce, while Amazon has adopted contractual Agent Policies and pursued litigation to restrict third-party agents." This contrast pairs a neutral-to-positive phrase "open protocols to enable" with a more negative framing "adopted contractual... and pursued litigation to restrict." The wording casts platform openness in a favorable light and Amazon’s steps as restrictive, which biases the reader toward valuing openness over control.

"A Ninth Circuit ruling for Amazon would strengthen platforms’ ability to block third-party AI agents under the Computer Fraud and Abuse Act. A ruling for Perplexity would limit the statute’s reach over locally run AI assistants and shift the fight to technical countermeasures and contractual rules." These hypothetical outcomes are presented symmetrically, but the first sentence uses "strengthen platforms’ ability to block" which might sound protective of platform power, while the second frames Perplexity’s win as "limit the statute’s reach" and "shift the fight," language that could downplay legal clarity and suggest messier outcomes. The choice of verbs nudges different emotional responses to each outcome.

"The outcome has broad implications for retail advertising models that rely on human visibility of sponsored listings and for the future interaction between autonomous shopping agents and e-commerce platforms." This statement centers the impact on "retail advertising models" and "human visibility," focusing on commercial consequences rather than consumer privacy or innovation benefits. That emphasis privileges business interests and frames the debate primarily as an economic/advertising issue.

The text expresses caution and conflict-driven tension through words about legal threats, injunctions, appeals, and litigation. Terms like "legal threat," "challenging," "barred," and "preliminary injunction" convey a sense of confrontation and urgency; this emotion is moderate to strong because the passage centers on court actions and the possible large consequences for a multibillion-dollar business. That tension shapes the reader’s reaction by signaling that a serious dispute is underway and that the outcome could matter a great deal, encouraging attention and concern. A related emotion of defensiveness appears in both parties’ described positions: Amazon’s contentions that the assistant "lets Perplexity bypass" revenue-generating features and that blocking agents is "necessary to preserve" billing integrity express a protective, assertive stance. This defensiveness is clear and fairly strong, intended to justify Amazon’s legal measures and to make readers see Amazon as protecting a legitimate business interest. Perplexity’s responses—phrases that stress user authorization, limited data transmission, and that the complaint concerns "lost advertising impressions rather than the kind of technological harm the statute addresses"—carry a tone of protest and exoneration. This blend of denial and mitigation is moderate in intensity and seeks to reduce readers’ suspicion, framing Perplexity as acting within user consent and technical limits. The passage also contains an undercurrent of uncertainty and doubt, notably when the district court "express[ed] doubt that the statute was a good fit" and when the Ninth Circuit "stayed that injunction pending expedited appeal." These expressions of doubt are mild to moderate but important; they signal legal ambiguity and invite the reader to see the case as unsettled rather than clear-cut. This uncertainty encourages readers to weigh both sides rather than accept one narrative. The description of stakes—Amazon’s "68.6 billion advertising business," "majority of Amazon’s ad revenue," and "broad implications for retail advertising models"—conveys a sense of gravity and consequence. That gravity is strong and intended to provoke concern about wider economic and industry impacts, making the dispute seem significant beyond the two companies involved. There is a policy and industry framing that carries a didactic, somewhat cautionary emotion when the text contrasts "open protocols" with Amazon’s "contractual Agent Policies and pursued litigation." This framing is moderate and aims to guide the reader into seeing this case as an example of two competing approaches to technology governance, prompting consideration of larger systemic choices. Finally, the passage hints at a future-oriented anxiety or anticipation about "who controls AI agents" and how the ruling would "limit the statute’s reach" or "strengthen platforms’ ability to block" agents; that anticipatory emotion is moderate and functions to push the reader to appreciate that the decision will shape future behavior, encouraging a forward-looking response. Together these emotions steer the reader toward taking the dispute seriously, recognizing competing legitimate claims, and understanding that the legal outcome could change industry practices. The writer uses emotionally charged legal and financial vocabulary instead of neutral phrasing to increase impact; repeating the core conflict—access versus control, user authorization versus platform integrity—reinforces the clash and keeps the reader focused on the central stakes. Comparisons between platform strategies, and the highlighting of large monetary figures, amplify the sense of importance and possible loss, making the situation seem more consequential than a narrow technical dispute. Those rhetorical choices raise urgency, frame the parties as opposing camps, and channel reader attention toward the broader implications of the case.