As sites rebuild for AI agents, rules on access and pricing will shape who benefits

Personal AI agents that book, shop and negotiate on someone’s behalf have gone mainstream quickly. MIT Technology Review reports Meta’s Muse hit 5 million downloads in its first month and agentic web traffic tripled from August to September, with most heading to e-commerce; OpenAI’s Dots and Meta’s Muse are among the apps credited with the surge (Muse downloads). The web’s response is mixed. Some sites block agents outright, Amazon is fighting Perplexity’s agents in court and has blocked Muse, and Yelp will not let AI book reservations; recent reporting found people struggling to use agents on sites from Pizza Hut to Yelp. Others, like Shopify, are building dedicated checkout lanes for agents.

Agents that contest a bogus medical bill or shop for cheaper car insurance are useful. The problem is that the web was not built for a visitor that is neither a person nor a classic bot, and businesses are now deciding how to treat it.

The bigger risk may not be blocking but agents welcomed on terms that favor platforms: prices that differ by whose agent asks, recommendations carrying undisclosed commissions, and a human-facing web that withers as money moves to agent lanes, plus constant prove-you’re-human checks for people browsing directly. That would make the internet worse even for people who never use an agent. Privacy concerns about agents compiling data on users’ contacts add to the stakes (Muse and your friends’ data).

Wharton’s Ethan Mollick expects that once it is common knowledge agents can chase refunds or haggle, customer service lines could be flooded so real people cannot get through, a plausible near-term harm that falls hardest on those without agents.

A judge recently ruled for Perplexity against Amazon, which has appealed, arguing agents threaten its advertising business because ads served to bots are hard to distinguish from ads seen by humans. That is a candid admission that the ad-funded web depends on human attention, and business-model defense may increasingly be framed as security. Some rules are easy, such as keeping card details out of agents’ hands, an approach American Express and others are pursuing. The contested questions matter more: can a business block a legitimate agent acting for a paying customer, or stop it from negotiating? Courts and platforms are setting answers while policymakers have largely not engaged. More in our AI agents coverage, regulation coverage and Meta coverage.