Agentic Advertising Management Protocols (AAMP)

Agentic Advertising Management Protocols (AAMP) is IAB Tech Lab's framework for redesigning digital advertising for an era in which AI agents - not humans - are the primary audience for commercial messages. Published in 2025, AAMP acknowledges a structural truth that the advertising industry has been slow to confront: when the customer sends an agent, the impression is dead.

Traditional digital advertising is built on a single assumption: a human being will see the ad. Every metric - impressions, click-through rates, viewability, attention time - presupposes a pair of human eyes. Every creative format - banner, video, native, interstitial - is designed to capture human attention. Every targeting strategy - demographic, behavioural, contextual - is built to reach a human decision-maker.

AAMP replaces this assumption with a new one: the entity encountering the commercial message may be an autonomous AI agent acting on behalf of a human. The agent does not see banners. It does not watch videos. It does not respond to emotional storytelling. It evaluates structured data, compares offers programmatically, and makes recommendations based on its principal's delegated criteria. Advertising, in the agentic age, must become machine-legible commercial information rather than human-directed persuasion.

The framework comprises three pillars, each addressing a different layer of the agentic advertising stack: research (understanding how agents work), protocols (standardising how agents communicate), and identity (verifying who agents are and what they are authorised to do).

AAMP is structured around three interdependent pillars that together form the infrastructure for agent-to-agent advertising.

Pillar 1: The Advertising Research Task Force (ARTF). The ARTF is the research arm of AAMP, tasked with studying how AI agents discover, evaluate, and recommend products and services - and how advertising must adapt to remain effective when the audience is a machine. The ARTF examines questions that traditional advertising research has never needed to ask: What data formats do agents prefer? How do agents weight different types of commercial information? What constitutes a "conversion" when the agent, not the human, makes the selection? The ARTF's research informs the other two pillars, ensuring that protocol design and registry architecture are grounded in empirical understanding of agent behaviour.

Pillar 2: Agentic Protocols (including AdCP). The second pillar defines the structured communication standards for agent-to-agent advertising transactions. The centrepiece is the Agent-Directed Commerce Protocol (AdCP) - a protocol that governs how AI agents negotiate, place, and measure advertising in machine-to-machine interactions. AdCP replaces the impression-based model with a protocol designed for autonomous decision-making. Under AdCP, advertising becomes structured data that agents can query, compare, and act upon without rendering visual creative. The protocol addresses four functions: discovery (how agents find relevant commercial offers), negotiation (how agents evaluate and select between competing offers), placement (how selected offers are integrated into agent recommendations), and measurement (how outcomes are attributed and verified).

Pillar 3: The Agent Registry. The third pillar provides the trust and identity infrastructure that makes agent-to-agent advertising trustworthy. The Agent Registry answers four questions about every agent participating in an advertising transaction: Who authorised this agent? What is it permitted to do? What is its track record? And can its claims be independently verified? Without a registry, every agent-to-agent interaction requires bespoke trust negotiation - making agentic advertising unscalable. The Agent Registry is analogous to certificate authorities in web security: it provides the chain of trust that enables strangers to transact.

The Agent-Directed Commerce Protocol (AdCP) is the operational heart of AAMP - the standard that defines how advertising actually works when both buyer and seller are represented by AI agents. AdCP is not a minor adaptation of programmatic advertising. It is a structural replacement for the impression-based model that has dominated digital advertising for two decades.

In traditional programmatic advertising, the transaction is simple: an advertiser bids for the right to show an ad to a human. The human sees the ad (or doesn't). The advertiser pays for the impression (or the click, or the conversion). Every element of this transaction assumes a human audience.

AdCP replaces this with a four-function protocol designed for machine-to-machine interaction:

Discovery. How does an agent find relevant commercial offers? In the human model, discovery happens through search, social media, or browsing. In the agent model, discovery happens through structured queries against machine-readable product catalogues. AdCP defines the query formats, response structures, and ranking signals that enable agents to discover relevant offers efficiently.

Negotiation. How does an agent evaluate and select between competing offers? Human negotiation involves persuasion, emotion, and brand affinity. Agent negotiation involves structured comparison of price, terms, quality signals, and alignment with the principal's delegated criteria. AdCP defines the negotiation protocol - how agents request customised offers, how sellers respond with structured proposals, and how agreements are formalised.

Placement. How are selected offers integrated into agent recommendations? In the human model, placement means showing an ad in a visible location. In the agent model, placement means integrating a commercial recommendation into the agent's response to its principal. AdCP defines how commercial content is distinguished from organic recommendations, how disclosure requirements are met, and how the principal's interests are protected.

Measurement. How are outcomes attributed and verified? In the human model, measurement means tracking impressions, clicks, and conversions. In the agent model, measurement means verifying that the agent's recommendation led to a transaction, that the transaction met the principal's criteria, and that the commercial arrangement was disclosed appropriately. AdCP defines the attribution model, the verification standards, and the audit trail requirements for agent-mediated advertising.

The Agent Registry is AAMP's answer to the Know Your Agent (KYA) problem - the fundamental challenge of establishing trust between entities that have never interacted before and may never interact again. In human advertising, trust is built through reputation, relationships, and regulatory oversight. In agent-to-agent advertising, trust must be established programmatically, instantly, and at scale.

The Agent Registry provides four capabilities:

Identity Verification. Every agent participating in an advertising transaction must be identifiable - not as an individual entity, but as a verified instance of a known agent class with documented capabilities and authorisation chains. The registry maintains cryptographic credentials that allow agents to prove their identity without revealing their principal's personal information.

Capability Declaration. Agents must declare what they are authorised to do. A shopping agent authorised to compare prices is different from a purchasing agent authorised to complete transactions. The registry maintains structured capability profiles that allow counterparties to understand what an agent can and cannot do before engaging in a transaction.

Behavioural History. Over time, agents accumulate track records - transaction completion rates, dispute frequencies, compliance adherence, and counterparty ratings. The registry maintains these histories as machine-readable reputation signals, enabling trust to be calibrated based on evidence rather than assumption. This infrastructure supports the broader shift toward zero-click commerce, where transactions complete without human interface interaction.

Authorisation Chain Verification. Every agent acts on behalf of a principal. The registry provides mechanisms for verifying the delegation chain - from human principal to agent to sub-agent - ensuring that every action can be traced back to a human who authorised it. This is the infrastructure-level implementation of what AXD calls trust architecture.

From the perspective of Agentic Experience Design (AXD), AAMP is significant not because it solves the advertising problem but because it reveals five structural lessons that apply to every sector facing the agentic transition.

Lesson 1: Attention is no longer the primary currency. AAMP's existence proves that the attention economy - the foundation of digital advertising for two decades - does not survive the agentic transition. When the audience is a machine, attention is meaningless. The new currency is structured legibility - the ability to be understood, evaluated, and acted upon by autonomous agents. This lesson applies far beyond advertising: every industry built on capturing human attention must redesign for machine evaluation.

Lesson 2: Trust requires infrastructure, not just intention. The Agent Registry is not a nice-to-have. It is the load-bearing element without which agent-to-agent advertising cannot function. This mirrors AXD's founding insight: trust architecture is not a feature - it is the primary material of agentic systems. Every sector building agent-to-agent capabilities will need its own version of the Agent Registry.

Lesson 3: Protocols precede products. AAMP is not a product. It is a protocol - a shared language that enables interoperability between agents from different providers. This is the same pattern seen in MCP, A2A, and ACP, in AP2, and in Google's Universal Commerce Protocol. The agentic economy is being built protocol-first, not product-first.

Lesson 4: The dual audience problem is universal. AAMP exists because advertising now serves two audiences simultaneously - humans who see ads and agents who evaluate commercial data. This is the same dual audience problem that AXD identifies across every digital surface: websites must serve both human browsers and AI agents, product catalogues must be both visually appealing and machine-readable, and trust signals must be both emotionally resonant and programmatically verifiable.

Lesson 5: Measurement must evolve from observation to verification. In human advertising, measurement means observing what happened (who saw the ad, who clicked, who converted). In agent advertising, measurement means verifying what happened - confirming that the agent's recommendation was appropriate, that the transaction met the principal's criteria, and that the commercial arrangement was disclosed. This shift from observation to verification is a core principle of AXD's approach to agent observability.

AAMP is the advertising industry's first structural response to the agentic transition - but its implications extend far beyond advertising. The framework reveals a pattern that will repeat across every sector: when autonomous agents become the primary interface between humans and institutions, every assumption built on human attention, human presence, and human decision-making must be redesigned.

The advertising industry's response is instructive because advertising was the first major digital industry built entirely on human attention. If attention-based advertising must be redesigned for agents, then attention-based everything must be redesigned. This includes customer experience (built on human navigation), brand strategy (built on human perception), pricing (built on human psychology), and loyalty programmes (built on human emotion).

AAMP also demonstrates the speed at which industry standards bodies are moving. IAB Tech Lab published AAMP within months of the first commercial AI agent deployments. This pace reflects the urgency of the transition - and the recognition that without shared protocols, the agentic economy will fragment into incompatible silos.

For organisations preparing for the agentic transition, AAMP offers a template: study how agents interact with your sector (research), define the protocols for agent-to-agent transactions (standards), and build the trust infrastructure that makes those transactions safe (registries). This is the same three-layer approach that the full AXD Observatory essay on AAMP explores in depth.

The age of the impression is ending. The age of the agent transaction is beginning. AAMP is the bridge between the two - and its three-pillar architecture will be replicated across every sector that the agentic economy touches.