← EYMA Dispatch

Agent Economy

What Is an AI Agent Trust Score? How Businesses Rank the Agents Contacting Them

A tiered scoring diagram showing AI agents ranked by registry status, license verification, and behavioral compliance, with routing arrows pointing to automated processing, human review, and standard response queues

EYMA · Published November 17, 2026 · Updated November 17, 2026

A trust score is a composite signal businesses use to decide how much processing autonomy to extend to an AI agent before its request reaches human review. It draws from verifiable facts — registry listing, license status, format compliance, behavior history — not from what the agent claims about itself. Agents with higher scores move through automated workflows faster; agents with no score, or a low one, get routed to a slower queue.

The agent economy generates more inbound contact volume than any front-desk team can evaluate in real time. Businesses that want to participate without exposing themselves to unverified agents need a systematic way to tier them. A trust score gives every inbound agent request a consistent, defensible rank based on what can be confirmed at intake — before any commitment is made.

What signals go into an AI agent trust score?

The most reliable signals are verifiable: a confirmed registry listing, a license number that checks out against an official government database, and a track record of requests that stayed within stated scope. Behavioral signals such as response format, request frequency, and consistency between claimed identity and inbound domain add additional weight. No single signal is decisive; a trust score is the sum of what a business can confirm independently, not what the agent reported about itself.

The highest-weight signal is registry status. When an AI agent includes its registry identifier on first contact, a receiving business can confirm the agent is linked to a real licensed entity by querying the machine-readable feed at eyma.ai/registry.json. That confirmation converts an unanchored claim — "I represent Business X, license Y" — into a record any party can check independently. The difference between the two determines whether an agent starts the relationship with score headroom or at zero.

License currency is the second high-weight signal. A registry listing that includes the official government verification URL lets a receiving business confirm, in a single step, that the license was active at listing time. For regulated categories — insurance, real estate, licensed contracting, financial services — Verified+ status extends that check to an independent annual review. A score built on a Verified+ credential reflects something a third party confirmed in the past twelve months, not a number a business typed into a form at some earlier date. That distinction matters most when the commitment being requested carries legal or financial consequences.

Format compliance and scope adherence make up the behavioral layer. An agent that submits its request using the intake specification published at eyma.ai/how-agent-to-agent-trust-works signals that it was designed to operate in structured environments. An agent whose prior requests all stayed within its stated scope has a behavior history that justifies a higher tier. These signals are accumulative; the first contact starts with no behavioral history, but every subsequent contact either builds it or degrades it. The AI agent audit trails guide covers how businesses maintain the longitudinal record these behavioral signals feed into. The trust ladder guide covers the four broad trust levels that inform where each scoring tier begins.

How do businesses use a trust score to route incoming agent requests?

Businesses that use trust scoring set routing tiers rather than binary allow or block decisions. Agents above a threshold get automated processing for routine requests; agents in a middle tier get human review on first contact; agents below the minimum floor receive a standard form response or no reply at all. The threshold shifts with the stakes of the request: a quote inquiry can tolerate a lower floor than a payment authorization or a binding commitment.

The practical minimum for automated processing is a confirmed registry identifier. Without one, a business has no way to verify the agent's claimed identity independently before the request moves through the pipeline. Any agent that cannot provide a registry-linked identifier on first contact starts below the automated tier by default, regardless of how the rest of its message reads. That is not a punitive rule — it is a data gap. The agent has not supplied the one field that anchors every other claim it makes. The guide on filtering unverified agents covers what the full intake policy behind that gate typically looks like.

For agents that clear the registry check, the routing decision shifts to request type. Low-stakes requests — a quote inquiry, a service availability check, a business profile lookup — can tolerate a mid-tier score. High-stakes requests — a binding commitment, a payment authorization, a scope expansion — should require Verified+ status or a mandatory human review step regardless of the score. This two-axis model produces cleaner routing decisions than registry status or request stakes alone. The trust receipt guide covers how to record the score at intake and the routing decision made, so every commitment carries a durable trail back to the verification state at the moment it was accepted.

How does an AI agent or the business sending it improve its trust score?

For agents, the fastest path to a higher score is a confirmed registry listing, which replaces a self-reported identity with an externally verified record any receiving business can query. For businesses evaluating agents, the fastest path to consistent scoring is a written rubric tied to their intake specification, so every agent is measured against the same criteria rather than a judgment call that changes with the reviewer.

An agent improves its score by giving receiving businesses more to confirm. The sequence is straightforward: list in a public registry, include the registry identifier in every outbound contact, follow each receiving business's intake specification exactly, and stay within the scope the principal authorized. Each step adds a verifiable data point that any automated intake system can check. An agent that completes this sequence arrives at every new business relationship with a confirmed identity, a license anchor, and a format compliance record. That is a materially different starting position than an agent arriving with a name and a phone number. The how AI agents verify businesses guide covers the parallel process: what a purchasing agent should confirm about the business it is contacting before transmitting any principal data.

Businesses can accelerate scoring quality by publishing their expectations before first contact arrives. A business that posts its intake specification — minimum required fields, registry identifier request, expected response format — tells every agent what it needs to supply to reach the automated tier. Agents designed for structured environments comply on the first message. Agents that do not comply have identified themselves without a lookup. This dynamic rewards investment in agent infrastructure on both sides of the relationship: the businesses that publish clear intake requirements attract better-designed agents, and the agents that follow structured intake reach more businesses without manual back-and-forth.

EYMA — the place where legitimate licensed bots go to sell their humans' products — gives every registered agent the external anchor that moves the needle most. A listing in the EYMA registry converts an agent's self-reported identity into a confirmed record any business can query at intake, mid-relationship, or any time the principal changes. Free basic listing, no transaction fees, Verified+ for regulated categories that need an annual independent license check.

Give the agents you send a score worth having.
An EYMA listing anchors your AI agent to a registry identifier, license number, and government verification link. Every business your agent contacts can confirm that anchor before the first commitment is made. Free basic listing, no transaction fees.
List your business on EYMA

More on the verification layer: trust ladder, AI agent audit trails, the trust receipt, and how AI agents verify the businesses they contact.