On November 4, 2025, Amazon issued a formal legal demand to AI startup Perplexity over Comet, a browser feature that lets users deploy autonomous shopping assistants on Amazon. The notice, the first reported legal threat by Amazon to an AI company, sparked immediate media coverage and a public rebuttal from Perplexity titled "Bullying Is Not Innovation." The exchange frames a larger debate about who controls AI driven shopping and how AI agents should operate inside major commerce platforms.

Why agentic shopping assistants matter

Agentic assistants are AI driven tools that can perform multi step tasks for users, such as searching for items, comparing prices, filling carts, and initiating purchases. Startups like Perplexity position Comet as an AI shopping agent that reduces friction, offers ecommerce personalization, and speeds decision making. For consumers the promise is convenience and tailored recommendations. For platforms this raises questions about transparency, site integrity, and compliance with terms of service.

Key claims in the dispute

• Amazon says Comet agents can act without proper transparency, degrade the shopping experience, and access the site in ways that violate its terms of service. Amazon framed the action as protecting customers and platform integrity.
• Perplexity responded publicly with "Bullying Is Not Innovation," arguing the agents run using users own stored credentials locally and therefore act as direct proxies for users rather than unauthorized bots. Perplexity warned the letter threatens the emerging market for AI agents.
• Coverage and commentary across multiple outlets highlighted the broader stakes for generative AI ecommerce and AI powered search, signaling that the outcome could set important legal and technical precedent.

Core technical and legal tension

The central question is whether agentic assistants are user controlled proxies operating with explicit user credentials and intent, or unauthorized automated actors that effectively scrape and transact on a platform without adequate disclosure. How regulators and courts interpret credentialed AI behavior will influence whether third party AI agents can operate freely on large commerce sites.

Implications for businesses consumers and the AI ecosystem

• Platform control versus user agency: Large platforms want to protect revenue user experience and security. Startups argue agents increase consumer choice and streamline commerce.
• Regulatory and technical precedent: This case could inform standards about credentialed AI acting on behalf of users and what transparency is required for AI agents.
• Market dynamics: If dominant platforms successfully limit agentic assistants innovation may concentrate inside platform controlled ecosystems. Allowing agents could spur competition in personalization and automation across commerce.
• Privacy and security: As AI agents handle sensitive credentials and personal preferences the need for clear privacy safeguards and secure local credential handling becomes central.
• Costs for startups: Legal pushback raises barriers to entry and could increase compliance costs for smaller companies building agentic features.

SEO and product guidance for companies building AI agents

To remain discoverable and resilient in an environment shaped by AI powered search and generative AI ecommerce platforms, integrate these best practices.

• Design transparency into agent workflows and surface consent clearly so users understand when an AI agent acts on their behalf.
• Use structured data and schema markup to improve visibility with AI agents and generative search systems. Provide product metadata price history and availability in machine readable formats.
• Optimize content for intent and AI driven discovery by using question format long tail phrases and topical clusters related to AI shopping agent functionality.
• Prioritize secure handling of user credentials and local processing when possible to reduce privacy risk and strengthen the argument that agents act as user proxies.
• Prepare for policy negotiations by auditing how agent automation interacts with platform terms and document consent flows and safety mitigations.

Conclusion

The Amazon versus Perplexity confrontation is an early but important test of norms for AI agents that act on behalf of users inside dominant online platforms. The outcome will matter for startups building agentic tools platforms defending their ecosystems and consumers seeking smoother automated shopping experiences. Companies should review how agentic automation intersects with platform terms adopt transparency best practices and optimize for AI powered search to stay competitive in the evolving landscape.