Guides

IAB Tech Lab's agentic advertising standards, explained: AAMP, ARTF, Agentic Audiences

IAB Tech Lab spent 2025 and 2026 building a stack of agentic advertising standards, and the names blur together fast: AAMP, ARTF, Agentic Audiences. This is the map. What each one is, how they relate, and the single strategic choice underneath all of them, which is to extend OpenRTB, AdCOM, VAST, and GPP rather than start over. For the rival stack, see the AdCP explainer; for how both touch the bid stream in practice, see agentic buying and OpenRTB.

The one idea underneath everything

IAB Tech Lab's agentic strategy starts from a position, not a product: the industry already has shared standards, and the fastest path to agents is to extend them. CEO Anthony Katsur has framed it as building "on an existing shared foundation, not introducing multiple new standards that create fragmentation." In practice that means OpenRTB stays the execution layer, AdCOM stays the object model, VAST stays the video envelope, and GPP stays the privacy transport. The agentic pieces sit on top and around those, adding negotiation, trust, and audience layers rather than replacing the bid stream. If you maintain OpenRTB integrations, that is the reassuring part: the protocol you already run is the thing they are building on.

AAMP: the umbrella

The Agentic Advertising Management Protocols (AAMP) reached v1.0 on January 28, 2026. It is the umbrella program, organized into three pillars:

  • Foundations. The shared vocabulary and object model agents use to describe inventory, audiences, and outcomes, anchored to AdCOM so an agent and an exchange mean the same thing by the same field.
  • Protocols. How agents talk: discovery, negotiation, and the handoff from a negotiated agreement to something that executes over existing rails such as deals in an OpenRTB bid request.
  • Trust. Authorization and provenance, so a buyer can verify that an agent is allowed to represent the inventory it is offering, extending the same supply-chain thinking behind ads.txt, sellers.json, and schain.

AAMP ships with buyer-agent and seller-agent SDKs published on GitHub, so the reference behavior is code, not just prose. The point of the three-pillar split is that a team can adopt the foundations and trust work without rewriting how auctions clear.

ARTF: agents inside the auction path

The Agentic RTB Framework (ARTF) is the most consequential piece for anyone who validates bid streams, because it is the one that touches the request in flight. Where most agentic activity happens at planning time, ARTF specifies a containerized execution model: a host, an exchange or a DSP, runs third-party agent services co-located in its own infrastructure, communicating over gRPC with Protocol Buffers, and those agents can enrich and modify a bid request mid-auction. Proponents claim large latency wins over server-to-server hops, because the agent runs next to the auction instead of across a network.

For OpenRTB, the implication is provenance. Under ARTF, the request that reaches a bidder may have been rewritten by a third-party container between the publisher and the auction. Which fields an agent may touch, and how that is audited, becomes as important as the transport itself. ARTF went through public comment closing January 15, 2026, and a version 2.0 is already underway, so the boundaries are still being drawn. The practical takeaway: if an ARTF-style stage sits in your path, validation stops being a one-time edge check and becomes something you want after each stage that can mutate the request.

Agentic Audiences: the targeting layer

Agentic Audiences reached v1.0 in November 2025. It is the audience-activation side of the program: a standard way for an agent to describe and request an audience so that targeting can be negotiated and executed without every buyer and seller inventing their own vocabulary. It sits alongside the existing data and segment fields in the bid request rather than replacing them, which keeps it consistent with the same extend-don't-replace philosophy.

Where OpenRTB sits in all of this

Across the three pieces, OpenRTB is not deprecated anywhere. AAMP anchors its vocabulary to AdCOM, which OpenRTB already uses for every enumerated list. ARTF rewrites bid requests but they remain OpenRTB bid requests on the wire. Agentic Audiences describes audiences that still travel in the request. The through-line is that agents change who decides and how deals are struck, and the decision still executes as a bid request that has to be valid. That is exactly why the roadmap this program came from spent most of its energy on agents while OpenRTB kept shipping ordinary dated snapshots; the two tracks are complementary, not competing. Our roadmap analysis covers the timeline in detail.

What to do now

  • Read the buyer and seller agent SDKs if you are prototyping. They are the fastest way to see what AAMP actually expects on the wire.
  • If you might host an ARTF-style agent, plan for request provenance: log what each stage changed, and validate the request after any stage that can rewrite it.
  • Keep your OpenRTB validation current regardless of which agent stack you adopt. The request still has to come out valid, so run it through the tester or the CLI.

Sources

rtblint is independent and not affiliated with IAB Tech Lab. Program names and structure below are drawn from IAB Tech Lab's own materials; the analysis is our own.