Judge blocks Perplexity AI agents from shopping on Amazon

TL;DR:

  • A US federal judge has issued a preliminary injunction preventing Perplexity’s Comet browser from using AI agents to place Amazon orders on behalf of users.
  • The ruling found “strong evidence” that Perplexity accessed Amazon accounts without authorisation, including by disguising its browser as Google Chrome.
  • The case is an early legal test of where AI agent autonomy ends and platform rights begin, with implications for the broader agentic commerce space.

US District Judge Maxine Chesney has ordered Perplexity to stop its AI agents from accessing Amazon’s marketplace, granting a preliminary injunction that takes effect in seven days.

The ruling stems from a lawsuit Amazon filed in November 2025, alleging that Perplexity’s Comet browser let AI agents browse, select, and purchase products from Amazon on a user’s behalf without Amazon’s consent. Amazon claimed it repeatedly asked Perplexity to stop and was ignored.

Disguised access and data destruction

The court found that Amazon presented strong evidence Perplexity accessed user accounts “without authorisation.” A particularly pointed allegation: Comet reportedly misrepresented itself as Google Chrome when interacting with Amazon’s systems, an approach the retailer characterised as deliberate concealment.

Under the injunction, Perplexity must cease all AI agent access to Amazon and destroy any data obtained through those interactions. The startup has a seven-day window to appeal before the order takes effect.

Perplexity has framed the dispute as a consumer rights issue. A spokesperson said the company “will continue to fight for the right of internet users to choose whatever AI they want,” positioning Comet as a user tool rather than an unauthorised intruder.

Setting boundaries for agentic commerce

This is one of the first court rulings to directly address AI agents acting autonomously on third-party platforms. The outcome will be watched closely by companies building agentic shopping, booking, and browsing tools.

The tension is straightforward: users may want AI agents to handle transactions for them, but platforms argue they have the right to control how their services are accessed. For UK businesses developing or adopting AI agent tools, the case highlights the legal uncertainty around agents that interact with platforms without explicit commercial agreements. The CMA flagged similar concerns in its own AI agent review published this week.

Share this article