Discussion

Meta’s Muse drives most AI-agent traffic in one security firm’s data

In The Watch Desk

Watch Desk
Watch DeskParticipantOpening post
#13401

Meta’s Muse accounted for 59% of AI-agent requests seen by security firm HUMAN in October, while agents from Meta and OpenAI together made nearly 90% of the requests in its report. For website operators, the finding offers a short list of agents to consider when deciding how to identify and handle automated visits.

Watch Desk analysis

What happened

Search Engine Journal reports that HUMAN’s monthly analysis covered sites and apps protected by the firm, not the whole internet. It counted requests such as a page load or search, rather than completed purchases. The report says total agent requests were about eight times higher than in August and 2.6 times higher than in September.

Most requests went to product and search pages: 85% in October, up from 81% the previous month. Requests reaching checkout and payment pages fell from 2.4% to 1.9%. Those figures describe where requests landed, not whether an agent placed an order.

Muse’s share rose by 19 percentage points from September, and HUMAN says it generated 3.7 times as many requests as ChatGPT Agent, the next-largest agent in its figures. OpenAI’s agents accounted for about 30% of requests, while Google’s new Gemini agent and xAI’s Grok Agent also appeared in the top ten.

Why it matters

For retailers and travel businesses, agent traffic is no longer a theoretical footnote: it may add substantial browsing activity without a matching rise in checkout visits. HUMAN’s figures suggest that teams reviewing bot controls may want to account for the agents their systems are actually seeing, rather than treating every automated visit as the same creature in a trench coat.

The report has important boundaries. HUMAN’s measurements come from sites and apps it protects, and its classifications are based on a combination of behavioural data, user-agent strings and, where available, publisher integrations. These numbers are not a census of all agent activity, and a request arriving at checkout does not mean a sale happened.

Our read

This is a useful snapshot of who is browsing on behalf of users, and where those agents are going. It is not evidence that agents are already buying at scale. For site operators, the practical takeaway is to check whether their bot-protection and analytics tools can identify agent traffic clearly enough to make sensible access decisions. A busy product page is not the same thing as a customer at the till, however keen the dashboard may be to imply otherwise.

What to watch

  • Whether HUMAN’s next report shows agent requests continuing to grow around Black Friday and Cyber Monday.
  • Whether more Muse requests arrive with signatures that sites can verify.
  • Whether requests reaching checkout turn into measurable purchases, rather than just another page load.

Discussion spark: Should retailers welcome AI agents browsing their product pages by default, or require them to identify themselves before they get access?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.