Perplexity AI says Amazon has warned the startup to stop its new AI agent from making purchases on the retail giant’s site. The notice landed Tuesday, according to the company, and targets the agent that runs inside Perplexity’s Comet browser. The move highlights a growing clash between automated assistants and big platforms that control online stores, data, and rules.
The warning centers on whether an AI agent can act like a person and shop on a user’s behalf. Perplexity positions the tool as a convenience feature. Amazon’s legal team appears to see a violation of its terms and a risk to its marketplace. The outcome could shape how future AI agents interact with major shopping sites.
“Perplexity AI said on Tuesday it has received a legal threat from Amazon.com demanding that the startup block the AI agent on its Comet browser from shopping on the ecommerce giant’s platform on a user’s behalf.”
Rising Tension Over AI Agents
AI agents can browse sites, compare prices, apply coupons, and check out. They promise faster shopping and fewer steps for users. They also test long-standing rules set by platforms to control automated access and transactions.
Amazon’s Conditions of Use restrict scraping, bots, and automated tools without permission. E-commerce companies cite fraud prevention, inventory protection, and fair use of their systems. Perplexity’s tool blurs the line between personal use and automated activity, since it connects to a store as if it were a human shopper, yet does so programmatically.
Similar conflicts have surfaced before. Ticketing sites have pursued anti-bot cases for years. LinkedIn fought third-party scraping of public profiles in a landmark legal battle. E-commerce has seen comparable pushback when software tries to automate cart activity, price tracking, or bulk purchases.
What Perplexity’s Tool Promises
Perplexity has pitched itself as an AI search and assistant company. Its Comet browser adds an agent layer that can perform tasks, not just return answers. Shopping flows are a clear use case, where a user might say “buy this item” and the agent completes checkout.
For end users, the appeal is speed and fewer errors. For sellers and platforms, the risk is load, misuse, and loss of control. Automated orders can stress anti-fraud systems or trigger pricing glitches. They can also bypass on-site features that platforms use to enforce rules or promote certain listings.
- Benefits for users: faster checkout, hands-free browsing, quick price checks.
- Concerns for platforms: bots, compliance, fraud exposure, and policy enforcement.
- Questions for regulators: consent, liability, and transparency of automated actions.
Legal and Policy Fault Lines
At issue is consent. Platforms argue they set the terms for how software interacts with their sites. Startups argue that users should be able to direct tools to act for them, like a password manager or accessibility tool would. The legal system has not fully resolved where AI agents fit in.
Courts have weighed in on scraping and automated access, often siding with platforms when terms are clear and technical blocks are in place. But consumer-directed automation is still a gray area. If a person can click through a checkout, can they authorize an agent to do the same using their account and payment details?
Security is a second front. Platforms can claim heightened risk from bots that mask behavior, even with a user’s blessing. Startups counter that agents can log actions, add confirmations, and reduce human mistakes. Neither claim has been comprehensively tested for AI-first shopping tools.
Industry Impact and Next Steps
The dispute could set a template for how AI browsing agents operate across the web. If Amazon’s position prevails, agents may need formal partnerships, certified APIs, or limits on actions. If Perplexity and similar firms find a path, consumer-led automation could expand to travel booking, subscriptions, returns, and warranty claims.
Standards may emerge to balance access with oversight. Possible measures include rate limits, verified agent identities, and audit trails for transactions. Clear consent prompts could help users understand when an agent is acting on their behalf and what data it touches.
For now, Perplexity faces a choice: disable shopping flows on Amazon, negotiate access, or risk a legal fight. Amazon, for its part, is signaling a strict stance on automation that touches its checkout. The decision will echo beyond one startup, affecting any company building agents that click and buy.
The latest twist shows how quickly AI features are colliding with entrenched platform rules. Users want convenience. Platforms want control and safety. Watch for whether the two sides land on technical guardrails and agreements, or head into court. The answer will shape the next generation of AI assistants and the limits of hands-free shopping.
