A Future Where AI Agents Shop? The Boundary of "Proxy Purchasing" Drawn by Amazon vs. Perplexity
In October 2025, a blog post titled "Bullying is not innovation" published by the AI search engine company Perplexity caused a stir. The trigger was Amazon sending a legal warning to Perplexity. The issue at hand was that Perplexity's AI shopping assistant, known as "Comet," when searching for or purchasing products on Amazon on behalf of users, "failed to clearly disclose that it is an AI agent." In this article, we will decode the "legal and ethical challenges of agentic AI" and the "new boundaries of platform dominance" that emerge through this incident.
1. The "Cease-and-Desist" Issued by Amazon—The First Legal Warning Regarding AI Agents
Amazon had warned Perplexity multiple times that "Comet is in violation of Amazon's terms of service." Ultimately, the company sent a formal "cease-and-desist" letter. This effectively became Amazon's first legal action against an AI company.
Perplexity argues the following in its blog:
"This week, we received aggressive legal threats from Amazon. They are demanding that we prohibit Comet users from using the AI assistant on Amazon. This is a threat to all internet users."
Perplexity's position is clear. It is the interpretation that "AI agents act as proxies for humans and have the same access rights as humans." Therefore, they argue that Comet has no obligation to disclose that it is an "AI" when acting on Amazon.
2. Amazon's Counterargument—"Proxy Actions Without Transparency Are Not Permitted"
Amazon clearly countered this.
"Other third-party agents, even when acting as proxies for users, identify themselves. Food delivery apps, travel agencies, and shipping apps are examples of this."
The company further added, "When third-party apps make purchases on behalf of users on other companies' sites, they should operate openly and respect the site operator's intent."
In other words, Amazon's position stands on the principle that "transparency is the prerequisite for trust."
However, behind this lies another realistic motive. Amazon also deploys its own AI shopping bot, "Rufus," and there is an underlying intention of economic defense to prevent third-party agents from intercepting its data and sales funnels.
3. Background: The Resurfacing "AI Bot vs. Website" Controversy
This incident reminds us of the conflict between Cloudflare and Perplexity that occurred in 2024. At the time, Cloudflare claimed that Perplexity was ignoring websites' "AI bot rejection settings" to scrape data. On the other hand, Perplexity countered that "when a user asks a question about a specific site, the act of browsing that site is the same as a normal browser."
As such, the line between "actions taken by AI as a proxy for humans" and "access by bots" is extremely ambiguous.
When AI "acts reflecting human intent," is it a proxy rather than a bot? Or, since it involves autonomous judgment, is it considered an independent entity?
Amazon's stance this time can be said to have set a new precedent in the industry by clearly positioning this line as "AI is an independent entity and requires the site's approval."
4. The Underlying Hegemony Struggle Over "Advertising and Sales Funnels"
Perplexity criticizes Amazon's motive as being the "maintenance of a monopoly on advertising and recommendations."
In a purchasing process where AI agents intervene, users become less susceptible to advertisements and promotions. For example, if Comet is ordered to find the "cheapest laundry basket," the AI will not add unnecessary items to the cart.
For Amazon, this means a loss of "impulse purchases."
This structure could potentially spread to Google's search advertising model and Meta's targeted advertising. If agentic AI becomes mainstream, the advertising economy itself could collapse in a "world where AI autonomously purchases and selects."
5. Future Outlook—Toward Rule-Making for the "Agent Economy"
The Perplexity incident is not merely a legal conflict, but a prelude to the era of the "agent economy."
In Silicon Valley, there is talk of the arrival of an "agentic world," where AI performs reservations, purchases, and negotiations on behalf of humans. However, Amazon's stance this time has sent a clear signal requiring "site operator approval and transparency" as a prerequisite.
If Amazon's argument becomes standardized, in the future, frameworks such as granting "digital identity (agent ID)" to AI agents or an "API registration system for access permission" will likely become necessary.
Conclusion
The clash between Perplexity and Amazon is not merely a technical dispute.
It poses an essential question: 'Who controls the internet?'
Are AI agents extensions of humans, or are they independent entities?
The battle over this ambiguous boundary will determine the rules of the web for the next decade.
'Transparency or freedom? It is not technology companies, but society as a whole, that must provide the answer.'