Encore Index Finds 22% of Organizations Block Declared AI Agents

Encore AI’s latest readiness scan found that 445 of 2,037 organizations block agents that identify themselves, measuring website access but not whether companies lose sales.

Completed Oct. 27, the scan covers organizations across 25 business types. It places 22% in the “Closed” category, where declared agents are blocked from public websites. Another 36 organizations, or 2%, are “Trusted,” meaning customers can grant agents limited access and revoke it later.

The index lists 114 organizations (6%) as “Connectable” through an MCP connection or public API program. Another 1,233 (61%) are “Legible,” meaning agents can read and cite their public content. These labels describe different kinds of access; they do not record completed transactions.

A separate October 2026 scan covered 8,638 sites and found that 21.6% blocked requests from agents that disclosed their identity. It used passive observation and did not simulate agent journeys, so its result is not directly comparable with Encore’s categories.

Encore’s methodology uses 24 checks of public websites and technical files, plus six calls to businesses’ published phone numbers. It does not sign in or assess pages behind a login. The index also does not track assistant recommendations, customer choices or lost revenue.

Dvir Ginzburg, Encore AI’s founder and CEO, said, “What gets rejected there is a label, not a capability.” He argues that an agent denied access may direct a customer to another provider, but the index does not measure that effect.

The scoring documentation has inconsistencies: some pages describe six scoring areas but list seven, while the FAQ’s weights differ from those in the research and methodology tables.