Retailers have spent decades refining the terms that govern what happens when consumers visit their websites, create accounts, make purchases, return products, and otherwise interact with their businesses online. Those terms were largely built around a familiar – and consistent – participant: a human user. Agentic commerce is complicating that model.
AI agents are starting to search for products, compare prices, navigate websites, and take actions on consumers’ behalf. And as more of the shopping journey stands to be carried out through an AI intermediary, retailers are faced with a deceptively basic question: Do the terms governing their digital operations still work when a machine participates in the customer journey?
The issue extends beyond terms that expressly address AI. Retailers rely on terms of use, conditions of sale, privacy policies, return policies, and other agreements and policies to govern purchases, accounts, acceptable use, dispute resolution, liability, and other aspects of the customer relationship. Many of those terms assume that the consumer is the one interacting with the retailer.
An AI intermediary changes that dynamic, potentially altering how terms are presented and assented to, what a consumer can authorize an agent to do, and which parties are bound by which obligations.
When Machines Encounter the Terms
Reddit’s case against Anthropic is already testing one part of this issue: whether a company can be bound by website terms when its automated systems access the site. In the lawsuit that it filed against Anthropic in San Francisco Superior Court in June 2025, Reddit alleges that the AI company repeatedly accessed and scraped its platform in violation of its User Agreement. The agreement restricts certain commercial and competitive uses and separately regulates automated access, including prohibiting scraping without Reddit’s prior written consent.
Anthropic argued in its defense that it was not bound by what it characterized as unenforceable browsewrap. A federal court rejected that argument in March, finding that Reddit and Anthropic were contractually bound based on Anthropic’s constructive and actual knowledge of the User Agreement.
In doing so, the court looked beyond whether an Anthropic employee clicked a box accepting Reddit’s terms. It pointed instead to evidence of Anthropic’s alleged knowledge of Reddit’s restrictions, including Reddit’s licensing discussions with Anthropic and Anthropic’s public commitment to honor robots.txt directives. Reddit also alleges that Anthropic continued accessing its platform through automated bots more than 100,000 times after publicly stating that it had stopped crawling Reddit.
After the case returned to state court, San Francisco Superior Court Judge Harold Kahn rejected Anthropic’s browsewrap argument in a September 17 order, allowing Reddit’s breach of contract claim to move forward. For companies seeking to regulate automated access through their terms, the rulings put the focus not only on what the terms prohibit, but on whether the company operating the automated system had notice of those restrictions.
The Agentic Commerce Question
Retail adds another layer of complexity to questions around AI-mediated access because an AI agent may be acting on a consumer’s behalf rather than accessing a site for its operator’s own purposes. That dynamic is at the center of Amazon’s case against Perplexity. As TFL recently reported, Amazon’s amended complaint adds a tortious interference claim to the case based on contracts between Amazon and its registered customers. Amazon alleges that its Conditions of Use, including its Agent Terms, govern customers’ deployment of AI agents and that Perplexity causes customers to violate those agreements when they use Comet – its AI-powered agent – on Amazon.
Amazon’s effort to enforce rules for AI agents through its customer agreements puts a different contractual relationship in play: the one between the retailer and the consumer using the agent. That relationship implicates what terms the consumer encountered, what the consumer authorized the agent to do, and which obligations bind the customer, the AI provider, or both.
IN PRACTICE: These are among the questions TFL will take up in our upcoming webinar, TRADEMARKS IN THE AGE OF AI, which will examine how AI is reshaping the legal framework for brands, including the implications for trademark and advertising law, scraping, and contracts, as well as what in-house and outside counsel can do now and how they should be preparing for what comes next.
Share
