Amazon fired the first volley with a cease-and-desist letter against Perplexity AI, accusing Perplexity of violating its terms of service and of potentially breaching federal and California computer fraud statutes.
Maya Mikhailov, CEO and founder at Savvi AI, told PYMNTS that Amazon’s move reflects a predictable defense of its revenue model. “Amazon’s reaction is very understandable,” Mikhailov said. “They’ve built this massive commerce and advertising engine, and if Perplexity or any AI browser starts to disintermediate them from their customers, that puts that entire stream at risk.”
Amazon’s letter demands that Perplexity “immediately cease using, enabling, or deploying Comet’s artificial intelligence agents” to access Amazon’s systems without permission. It alleges that Perplexity’s Comet browser scrapes Amazon data, misrepresents delivery times, and undermines product security and customer trust. The letter warns that Comet’s behavior could “erode customer confidence” and expose users to “phishing, scams, and other cyberattacks.”
In response, Perplexity rejected those allegations in a blog post Bullying Is Not Innovation, describing Amazon’s action as an attempt to block independent AI tools from operating across the open web. The company wrote that “large corporations use legal threats and intimidation to block innovation,” arguing that Comet acts only on behalf of the user. According to Perplexity, all credentials remain stored locally, and its agent “works for the user, not for Perplexity, and certainly not for Amazon.”
Data Access and Platform Control
It’s an understatement to say there’s a lot of money at stake. Amazon’s online advertising business grew 24% in the third quarter to $17.7 billion, outpacing the company’s AWS cloud computing unit, which posted 20% sales growth, according to the company’s earnings. Comet threatens that model by collapsing the multistage shopping process of discovery, comparison and purchase into a single action, skipping the sponsored results and recommendations that drive Amazon’s profit.
However, Amir Sarhangi, founder of Skyfire, told PYMNTS that the deeper issue is data visibility. “Nobody wants to get disintermediated,” he said. “For Amazon, that’s the real problem. They’ve always had visibility into what people browse, abandon and return to. With agentic shopping, they lose that view until checkout. By then, it’s too late.”
Echoing that sentiment, Entersekt Chief Technology Officer Gerhard Oosthuizen told PYMNTS, “The conversation around agentic commerce touches so many ecosystems, including brand loyalty, fraud, consumer trust and revenue disintermediation. But for eCommerce, the conversation largely stems around the reduction in information merchants might receive around client profile, interest, location and demographics.”
Data Wars and New Ways to Verify
Mikhailov said the case mirrors earlier fights over data access and pointed to a dispute over the summer between JPMorgan and Plaid over the FinTech’s access to consumer data. “This is completely analogous to JPMorgan and Plaid. Nobody wants a free ride on the back of a multibillion-dollar brand they built. In this case, it’s not bank data being used, it’s merchant and product data.”
Oosthuizen added that the challenge also extends to how merchants manage bot traffic. “Merchants and financial institutions have long had the tools to detect and block bots from accessing their systems,” he said. “These malicious bots typically rely on scraping, credential stuffing or man-in-the-middle attacks. We can expect similar scrutiny as unapproved or undeclared AI agents begin interacting with merchant platforms. If they don’t properly identify themselves or are seen to be disintermediating the brand, they’ll likely be blocked.”
Sarhangi believes the lack of standards fuels this conflict. “There’s no good way right now for merchants to know who’s behind the agent. They don’t know who’s shopping or what’s being browsed. That’s why verification matters. Knowing your agent changes everything.”
His proposed Know Your Agent framework would verify both the user and the AI system. “With KYA, the merchant can see who the consumer is, who the agent developer is, and what permission has been granted. That restores transparency and keeps the customer relationship intact.”
Next Gateway: Agents vs Platforms
McKinsey estimates that agentic commerce could reach $5 trillion in global transactions by 2030. OpenAI’s Instant Checkout already lets users complete purchases inside ChatGPT, and Walmart’s AI-first shopping integration offers a similar functionality to millions of customers.
“In the long run, this is about the operating system of consumer life,” Mikhailov said. “These assistants are trying to be everywhere across devices, voice and wearables. Whoever controls that agent layer controls the next digital gateway.”
For Amazon, the greater risk is losing visibility into consumer behavior that drives its advertising and recommendations. “Advertising will change, but it won’t disappear,” Sarhangi said. “Merchants still want to know who you are so they can suggest what goes with what. If the agent cuts them out of that loop, the economics collapse.”
He added, “In the end, this is a fight over identity. If Amazon can’t tell who its customer is, it can’t personalize, advertise or build loyalty. That’s the heart of this battle.”
Mikhailov said that while startups see opportunity, the industry still needs structure. “As a founder, I want freedom to build, but as someone who’s worked in banking and payments, I understand the need for guardrails. It shouldn’t take a major fraud incident for the industry to set standards.”
