THE CRUNCH
AI shopping agents from Meta, Google and OpenAI promise effortless buying, but in practice their access to retailers is patchy and sometimes blocked outright. A hands-on test of Meta's Muse agent found it caught by Target's bot checks one day and admitted the next, suggesting listings that would not ship to the tester's address, and unable to browse Amazon at all. Amazon has blocked Muse from its site and previously sued Perplexity, saying the AI company ignored requests to stop its agent shopping there.
The blocks extend beyond retail. JPMorgan Chase CEO Jamie Dimon said the bank would not allow outside agents to access customer accounts, calling it too risky, and some users report agents failing to book trips with providers such as Delta and Amtrak. General bot-blocking firewalls can also snarl agents even where no blanket ban exists, making results inconsistent.
Amazon's position has a twist: it is not against automated shopping in principle. The company runs its own on-platform agentic feature that monitors prices, reorders items and buys products based on user-set rules, plus a tool that purchases from other sites without leaving Amazon, which smaller businesses say they never consented to. An Amazon spokesperson confirmed the company asked Meta to remove Amazon from Muse, saying third-party purchasing apps should operate openly and respect service provider decisions.
The practical result, per the testing, is that directing an agent, waiting for responses and correcting its mistakes could take longer than searching yourself. Muse itself acknowledged the gap: when asked for the best per-piece granola bar price, it noted Amazon might have the lowest option but could not confirm because it was hard-blocked.


