Overview

Effective governance of agentic traffic starts with knowing which AI agent is behind each request. DataDome's agent identification engine determines the identity of every AI agent sending traffic to your platform in real time, using a layered approach that combines strong cryptographic methods, authoritative sources, and advanced fingerprinting techniques.

Every identified agent is assigned a transparent identification strength, so you always know how reliable a given identification is.

Identification Methods

DataDome applies a combination of identification techniques of varying strength:

Identification Method	Strength
Web Bot Authentication	High
Cryptographic token verification	High
Trusted IP source lists	High
Reverse-DNS (rDNS)	High
Network attribution (isolated AS or IP ranges)	Medium
Client-side fingerprinting	Medium
Server-side fingerprinting	Low

Agent identification is powered by two complementary mechanisms:

The DataDome Agent Catalog

DataDome takes full ownership of building, validating, and maintaining a comprehensive catalog of AI agent detection models. The research team actively monitors industry trends and traffic patterns to ensure the catalog remains current as the agentic ecosystem evolves.

Detection models in the catalog rely on the full spectrum of identification methods above, from high-confidence signals such as Trusted IPs and rDNS, to network attribution and fingerprinting for agents that cannot be identified through stronger means.

How new agents are added to the catalog:

Research and Modeling: DataDome's research team conducts a full technical and business assessment of the agent, and builds a detection model initially deployed in simulation mode.
Customer Notification: Once validated, the model is made visible to customers in a deactivated state. Customers are proactively notified and given a two-week window to review and, if needed, take action. Each new model is presented with a description, a Trust Score, and a DataDome-recommended policy.
Activation: After two weeks, the model is activated and the recommended policy is automatically applied for any customer who has not configured a custom policy.

This process ensures that every agent in the catalog has been individually assessed and validated by DataDome, providing customers with a trusted, curated foundation.

Third-Party Trust Authorities

Some AI agents operate under the certification of third-party trust authorities, such as Visa or Skyfire, that authenticate agents using emerging industry protocols like TAP (Token Agent Protocol) or KYA (Know Your Agent).

DataDome partners with the leading authorities in this space:

DataDome ensures real-time identity verification at the point of the transaction.
Third-party authorities provide independent assessment and validation of agent legitimacy and business relevance.

Customers automatically benefit from the agent catalogs of the major authorities, with no additional configuration required.

AI Agent Categories

DataDome classifies identified AI agents into four categories, reflecting the nature of their activity:

Category	Description
AI Crawler	Automated bot that systematically browses and indexes web content using AI to understand context and relevance.
AI Assistant	Interactive software that uses natural language processing to answer questions, perform tasks, and help users.
Autonomous Agent	AI system capable of setting goals and executing complex tasks independently without constant human intervention.
Agentic Browser	Web browser integrated with AI agents that can actively navigate, interact with, and perform tasks on websites.

Identification Strength

Every identified agent is assigned an identification strength reflecting how the identification was established:

Strength	Methods
High	Web Bot Authentication, cryptographic token verification, trusted IP source lists, reverse-DNS
Medium	Network attribution, client-side fingerprinting
Low	Server-side fingerprinting

Identification strength is surfaced in the Access Control and Explore dashboards, and is one of the key inputs to the AI Agent Trust Score.

Agent Catalog

DataDome maintains a continuously expanding and updated catalog of known AI agents, covering agents from major AI labs, search engines, data providers, coding assistants, autonomous agent frameworks, and other sources.

When a new AI agent is observed in the wild, DataDome's research team evaluates it and adds it to the catalog with the appropriate identification method and metadata. Existing entries are updated as agent operators change their infrastructure, rotate IP ranges, or adopt new identification protocols.