AI agents are showing up in more places than most people have had time to think carefully about. They handle transactions, interact with other software systems, and in some cases operate without anyone watching in real time. That expansion raises a question that sounds simple but turns out to be genuinely difficult. When an agent contacts your system, how do you know it is who it claims to be?
GoDaddy and LegalZoom are attempting to address that question through a new initiative built around an open standard called Agent Name Service, or ANS. The approach draws on infrastructure the web already uses: domain names, DNS records, and cryptographic certificates. Rather than building a parallel identity system from scratch, GoDaddy anchors agent identity to existing verification mechanisms that businesses and developers already understand.
The concept works by assigning each AI agent a unique, human-readable name tied to a verifiable domain. When another system encounters that agent, it runs a lookup to confirm the agent’s origin and checks whether it connects to a legitimate, identifiable organization.
LegalZoom has already registered an agent under this system, linking it to the company’s domain and making it discoverable by AI assistants and other systems that support the Model Context Protocol. That agent handles document workflows and facilitates access to legal services through verified identity rather than an assumed one.
The timing matters. Right now, an agent can appear functional, respond appropriately, and still offer no clear path back to whoever deployed it. In finance, legal services, healthcare, and other sectors where accountability ties directly to outcomes, that ambiguity creates real exposure. ANS attempts to make identity queryable at the agent level, giving systems a way to check origin before they accept instructions or data.
The honest limitation of the approach is that domain ownership and cryptographic proof confirm where an agent came from, not how it will behave. A verified agent can still carry misconfigurations, operate outside intended parameters, or take actions its owner did not anticipate. Identity forms a necessary layer, not a complete answer.
Adoption will determine whether ANS becomes a foundational part of how AI agents operate on the web or remains a standard that some providers use and others bypass in favor of proprietary alternatives.
