Agentic Ad Buying Gets Standards: What ARTF and AdCP Mean for Sellers

Two standards in one month

For most of 2025, agentic advertising was a phrase in conference keynotes: AI agents that plan, negotiate and buy media on behalf of advertisers, and sell it on behalf of publishers. This fall, it gained specifications. On October 15, a group of companies launched the Ad Context Protocol (AdCP). On November 13, IAB Tech Lab released the Agentic RTB Framework (ARTF) v1.0 for public comment. They address different parts of the problem, and publishers should understand both.

ARTF: agents inside the auction

IAB Tech Lab's ARTF announcement describes a container-based architecture that lets services, including AI agents, run inside a host platform's environment, such as the same data center, server or virtual machine as an exchange, instead of being called across the network. It defines how those containers behave and how they can interact with bid requests in real time, and it supports agent communication frameworks such as the Model Context Protocol (MCP) and Agent-to-Agent (A2A).

The pitch is speed. IAB Tech Lab says the approach can reduce bid request and response time by up to 80%, which would leave more of each auction's time budget for enrichment and decisioning. Index Exchange, Chalice and The Trade Desk were among the companies quoted in support. The public comment period runs until January 15, 2026.

In plain terms, ARTF is about the real-time plumbing: how a buyer's or seller's logic can sit close to the auction and act on individual impressions without adding network round trips.

AdCP: agents that plan and buy

The Ad Context Protocol works at a different layer. It is an open protocol for AI agents to discover inventory, plan, buy and measure media across platforms. It builds on MCP, so a buying agent can query a seller's agent about available products, pricing and audiences, then set up and manage campaigns through a common interface rather than a custom integration for every platform. Founding members at launch included Scope3, Yahoo, PubMatic, Triton Digital, Optable and Swivel, and the effort is led by Brian O'Kelley, who founded AppNexus.

AdCP is closer to how direct and programmatic guaranteed deals are negotiated today: finding the right inventory, agreeing on terms and activating a campaign. ARTF is closer to how open auctions run.

How they relate

It is tempting to frame these as competing standards. They overlap less than the headlines suggest. One is about workflow and negotiation between agents; the other is about running logic inside the real-time bidstream. What matters for publishers is that both assume the same thing: software will increasingly make buying decisions, and it will make them based on what it can read about your inventory.

What changes for sellers

Your inventory needs to be machine-readable

An agent choosing between publishers cannot read your media kit or take a call with your sales team. It will rely on structured information: what formats you offer, audience and contextual signals, viewability and attention history, content categories, pricing and deal terms, and supply chain transparency. Publishers whose inventory is described clearly and consistently in these terms will be easier for agents to buy.

Supply chain trust becomes automated

Agents will lean on the same trust signals buyers use today, such as ads.txt, sellers.json and supply chain objects, but they will apply them without exception. Inconsistent or stale entries that a human buyer might overlook could simply remove your inventory from an agent's consideration.

Pricing control matters more

If buying agents can query many sellers quickly and negotiate at scale, sellers need clear rules for what they will accept. Floors, deal terms and approval rules should be deliberate, documented and enforced in your systems, not left to case-by-case judgment that an automated counterpart will never see.

Questions still open

Both efforts are early, and several questions matter to sellers. How will a publisher decide which agents are allowed to transact on its inventory? How will fees for agent services be disclosed along the supply chain? What audit trail will exist when an automated buy goes wrong? Watch how the standards answer these during comment periods and early implementations.

What publishers should do now

1. Audit your inventory metadata. Check that content categories, formats, sizes and page-level signals are passed consistently in bid requests and described accurately in your deal and product listings.
2. Clean supply chain files. Review ads.txt, app-ads.txt and your sellers.json entries with each partner.
3. Ask partners about their plans. Ask your SSPs whether they plan to support ARTF containers or expose AdCP interfaces, and what control publishers will have over which agents can transact on their inventory.
4. Define your rules. Write down minimum prices, category exclusions and approval requirements for automated buys, so they can be encoded in whichever systems you use.
5. Comment if you care. ARTF is open for public comment. Publishers with engineering resources can influence how seller controls are specified.

The takeaway

Agentic advertising is moving from concept to specification. The standards are early, adoption will be uneven, and most publishers do not need to build anything this quarter. What they can do is make their inventory legible to machines, keep their supply chain signals clean, and turn pricing and approval rules into explicit policy. Those steps pay off today with human buyers and position you for automated ones. HBDR works with publishers on exactly that groundwork: clean signals, clean supply paths and pricing rules that hold up regardless of who, or what, is buying.