Personal AI assistants such as Meta’s Muse, Instinct, and ChatGPT’s Dots are moving beyond simple question-answering. They can complete end-to-end tasks for users, including reserving restaurant tables, purchasing groceries, and arranging travel, all without requiring the consumer to install or configure complex software on a personal computer.
Consumers report that many sites interrupt these workflows by denying the agents access. Amazon, for example, recently stopped Muse from browsing or buying items on its retail platform. Other complaints point to similar denials that stem from generic anti-bot defenses rather than deliberate policies, leaving users uncertain whether the block is accidental or intentional.
Walmart says its recent partnership with Muse is meant to meet shoppers where they are, and it does not view the restrictions as intentional. The retailer’s site employs a click-to-verify button that confirms a human visitor; when an agent cannot complete that step, the session is terminated. This highlights how existing captcha solutions may be ill-suited for an agent-first internet.
To mitigate the friction, a coalition that includes Meta, Walmart, Stripe, Sierra, Genesys, Rocket, NiCE and Decagon is drafting an open protocol for agent-to-business communication. The specification is intended to differentiate legitimate personal assistants from malicious bots in the context of online commerce, and Meta announced the effort in a recent blog entry.
Not all brands are ready to open their platforms. Delta’s spokesperson explained that the airline is safeguarding customers from unauthorized automated activity and is only evaluating how external agents might fit its digital experience. Yelp requires agents to purchase data-licensing access before any non-human traffic can retrieve listings, while eBay’s policy bars unauthorized scraping and model training but does not forbid all third-party assistants. Adidas, Zillow and Pizza Hut declined to comment, yet users have cited blocks on those sites as well as on various airlines.
Cloudflare, a major content-delivery network, offers tools that can block AI-driven traffic. A September 15 change to its crawler defaults lets site owners deny AI training bots while still allowing other crawlers, a shift some link to recent Muse disruptions. Cloudflare said it lacks specific data on the incidents and pointed to its public Radar service, which tracks overall bot volume but does not separate benign from harmful agents. The firm also runs a marketplace where AI bots pay for data access and is testing a browser designed for agents.