Jefferies finds Meta’s Muse AI agent blocked on about one-third of websites, raising questions on agentic commerce rollout

Meta Platforms’ AI agent Muse is running into practical access limits on a meaningful portion of the web, according to a market report citing an analysis by Jefferies.

The Jefferies work, described by the report, tested websites for whether Muse could access them and found that roughly one-third were blocked. The finding points to a broader distribution challenge for AI agents: even when users can ask for tasks, the agent still needs permission and technical access to complete them across third-party sites.

While the report did not specify which categories of websites were most affected, it framed the access issue as potentially delaying revenue tied to agentic commerce. Agentic commerce generally refers to AI systems autonomously taking steps toward buying or completing transactions, such as comparing products, navigating checkout flows, or initiating orders on a user’s behalf.

Jefferies’ conclusion, as relayed, suggests the monetization timeline for Muse could slip if the agent cannot reach enough destination websites to reliably execute commerce workflows. The concern is less about whether Muse can generate answers and more about whether it can consistently carry out downstream actions that depend on external site availability and permissions.

For Meta, the stakes are higher because Muse sits in the company’s push to expand beyond traditional chat-style assistants toward AI that can do tasks. In that model, performance bottlenecks like blocked access can translate into fewer successful commerce attempts, less user trust, and potentially reduced advertiser or commerce partners’ interest if automation rates lag expectations.

Meta has discussed its AI ambitions through product and technology updates, including ongoing work around AI agents and the infrastructure needed to support them. However, the market report did not cite Meta offering additional details on how Muse handles website restrictions, what partner sites are most critical, or whether it has mechanisms to route around blocks.

Notably, the report’s framing is cautious on timing. Even if access constraints limit early transactions, the path to improvement could involve closer integration with participating merchants, changes to how the agent interacts with websites, or evolving policies from the targeted sites. The report itself, as presented here, did not provide evidence on whether Meta can remediate the blocking problem quickly or what portion of commerce value sits on accessible websites versus blocked ones.

It is also unclear how materially the “one-third blocked” measure translates into customer experience and revenue. The test results, as described, are directional and do not quantify commerce success rates, user drop-off, or revenue per attempted transaction. Without those benchmarks, the practical financial impact remains a question for investors and analysts to monitor.