Websites Block AI Agents as 5 Million Meta Muse Downloads Push Internet Redesign
Updated
Updated · MIT Technology Review · Oct 19
Websites Block AI Agents as 5 Million Meta Muse Downloads Push Internet Redesign
3 articles · Updated · MIT Technology Review · Oct 19
Summary
Agentic web traffic tripled from August to September, and many sites are now blocking personal AI agents sent to book travel, compare prices or make reservations.
Meta’s Muse was downloaded 5 million times in its first month, accelerating a new class of web visitor that acts for individual users but moves at machine speed and does not fit existing bot-versus-human rules.
Amazon has blocked Muse’s agents and is appealing a ruling favoring Perplexity, arguing agents threaten its advertising business by blurring whether ads are seen by humans or bots; Yelp also bars AI restaurant bookings.
Shopify is moving the other way with agent-specific checkout lanes, underscoring how e-commerce incentives could reshape the web around AI shoppers rather than human browsing.
The broader fight is over rules still being written: whether businesses can block agents or stop them negotiating on users’ behalf may determine if the web becomes more useful for consumers or more hostile to humans.
As websites begin charging AI agents for access, will your personal assistant secretly make the internet more expensive for you?
With courts now protecting AI web browsing, are we witnessing the end of human-focused advertising and traditional websites?
If the internet optimizes for machines over humans, will everyday users soon find themselves locked out of the web's best deals?
$385 Billion at Stake: The Rise of Agentic AI Commerce and the Web’s Security Backlash in 2026
Overview
Meta Muse’s launch in September 2026 sparked a record-breaking surge in downloads, fueled by Meta’s aggressive promotion and deep platform integration. This rapid adoption quickly drew backlash from major web platforms like Amazon, which blocked Muse for bypassing ad revenue streams and operating without notice. As security providers like Cloudflare enabled default AI blocking, automated defenses spread across the web, causing both agent failures and increased friction for human users. Meanwhile, banks and financial institutions barred external AI agents over fears of fraud and liability, while legal frameworks lagged behind. The resulting standoff highlights the growing divide between open agentic innovation and entrenched platform controls.