Meta and OpenAI Agents Drive 90% of Measured AI Traffic

AI agents from just two companies — Meta and OpenAI — generated nearly 90% of all agent requests hitting sites protected by HUMAN Security in October 2026. The figure comes from the security vendor's monthly State of Agentic Traffic report and offers a rare concrete measurement of agentic AI actually running in production, rather than another adoption survey.

The concentration is stark. Meta's Muse agent alone produced 59% of all agent requests in October, up 19 percentage points from September, with request volume nearly quadrupling month over month. OpenAI's three agents — ChatGPT Agent, Codex Browser and the Codex Chrome extension — together accounted for about 30%. ChatGPT Agent's share fell from 28.5% to 15.9% even as its own volume grew 44%; Muse simply grew faster.

Total agent requests on HUMAN-protected properties ran roughly eight times higher than in August and 2.6 times higher than in September. The surge tracks Muse's launch on September 8 and its continued growth through October. Two platforms, in short, now dominate the agent web that HUMAN can see.

What does the data actually measure?

HUMAN's numbers come from its Sightline Cyberfraud Defense across applications on the Human Defense Platform. The company identifies agents through behavioral signal analysis, user-agent attribution, publisher-level integration and its AgenticTrust system, which uses proprietary research and digital signature validation. Crawler traffic is excluded.

Each request represents one agent action — a page load or a search — not a user or a session. The data covers only HUMAN's customer base, not the internet at large. HUMAN itself concedes the limit: on most sites, analytics tools cannot distinguish AI agents from human visitors, let alone identify specific agents or classify their intent. Any generalization from this sample to the whole web is the reader's inference, not the vendor's claim.

Are standalone agent browsers dying?

The evidence says yes, for consumers. Perplexity's Comet browser, which led with 41% of agent requests in August, fell to 3.6% in October — the only top-10 agent whose volume declined, dropping about 30% while total traffic grew eightfold. Genspark's agentic browser dropped out of the top 10 entirely. OpenAI discontinued its standalone ChatGPT Atlas browser in August.

HUMAN attributes the slide to users choosing agents embedded in platforms they already use, from companies that already reach them at scale. Developer tools tell a different story: OpenAI's Codex Browser nearly tripled its traffic in October. Convenience, not capability, appears to decide the consumer market.

Who else is showing up?

New entrants remain marginal. Google's Gemini agent, launched October 8, entered the rankings in eighth place with 0.8%. xAI's Grok Agent, launched August 11, debuted ninth at 0.7%.

A quieter development concerns verification. Between November 7 and 9, Muse moved its Web Bot Auth signatures from Meta's key directory (meta.com) to its own (muse.ai). As of November 10, roughly half of Muse's requests on DataDome's network carried no signatures at all — sites verifying against meta.com would now flag Muse as unverified. Both DataDome and HUMAN sell bot identification services, a commercial interest worth noting when evaluating their framing.

Do agents actually buy anything?

Mostly not yet. In October, 85% of agent requests went to product and search pages, up from 81% in September. Requests reaching checkout and payment pages fell to 1.9% of agent traffic from 2.4%, though absolute checkout volume grew 117% month over month. HUMAN cautions that a checkout-page request does not mean an order was placed.

Ecommerce sites held a stable 60% share of agent requests. Travel took second place for the first time at 19.3%, up from 17.1%, narrowly beating media at 19.1%. SaaS, at just 1.2% of traffic, grew fastest — nearly quadrupling in a month and running about 15 times its August level.

How does this compare with enterprise surveys?

A separate Zapier survey of 513 U.S. enterprise leaders at organizations with 1,000 or more employees, fielded September 16–21, found 88% say their organizations use or test AI agents, up from 72% in 2025. Production use jumped from 40% to 71%. But only 30% of leaders at agent-using organizations say the technology handles a majority of tasks without human intervention. Ninety percent report at least one positive outcome, and 46% say their most mature agent has paid for itself or is expected to within 12 months.

The two sources measure different things: Zapier captures self-reported internal adoption; HUMAN captures observed external traffic. Neither validates the other — intentions and observable web activity can diverge sharply.

For operators, the practical takeaway is narrow: HUMAN's report reduces the agent threat-and-opportunity surface to two vendors. The company advises ecommerce and travel teams heading into Black Friday to raise Muse and OpenAI's three agents with whoever runs their bot protection, and to point signature checks at muse.ai — while acknowledging that about half of Muse's requests currently carry no signature to check. Next month's report will show how agents behaved over Black Friday and Cyber Monday.