AdCP, One Year In: What Agentic Ad Buying Has Changed

The Ad Context Protocol did not stall. Almost a year after its launch in October 2025 by a group including Scope3, Yahoo, PubMatic, Optable, Swivel and Triton Digital, AdCP has a public specification on its third major version, a reference implementation on GitHub, and an industry body, AgenticAdvertising.org, that stewards it. Its scope has grown from agents negotiating deals to covering discovery, planning, creative preparation, audience activation, governance and reporting, plus a serve-time profile for activating pre-negotiated packages at the moment an impression is served.

What it has not done is kill programmatic’s black box. A year in, the more accurate description is that AdCP changed who sits in the middle of an ad deal and how they talk, and left the question of who can see the money largely where it was.

What has actually run on it

The live examples are real and still small. In December 2025, agency Butler/Till and PubMatic ran what is widely described as the first fully agentic campaign over AdCP, for the drinks brand Clubtails on connected TV: a natural language brief went in, and PubMatic’s sell-side agents handled planning, setup and optimization, as Star’s overview of agentic standards summarizes.

In January 2026, NBCUniversal, FreeWheel, Newton Research and the agency RPA announced an agent-to-agent buy across linear TV and streaming, including live football playoff inventory. The buy-side agent sends campaign goals to sell-side agents, they refine a proposal through automated feedback, and it executes once humans on both sides approve, according to AdExchanger. One precision point: NBCUniversal’s announcement describes the transaction as running on the Model Context Protocol, the general standard AdCP is built on, and does not name AdCP itself. Agentic buying and AdCP are related, and they are not the same thing.

That distinction matters because AdCP is not the only standards effort. IAB Tech Lab released an agentic roadmap in January 2026 that builds on existing standards such as OpenRTB rather than replacing them, and in February 2026 grouped its agentic work under a program it calls AAMP, per Star’s overview. So the industry is running two tracks at once: a new protocol designed for agents, and the incumbent standards body extending what already exists.

What “black box” meant, and what AdCP touches

The black box complaint about programmatic was never one problem. It was at least three: advertisers couldn’t see how many intermediaries touched a buy, couldn’t see what each one took, and couldn’t see how the same audience was being sold more than once along the way.

AdCP addresses the first most directly. Instead of an impression passing through a chain of exchanges and platforms, a buyer’s agent and a seller’s agent exchange a brief and a proposal in a common format, and the deal can be written straight into the seller’s systems. Fewer hops, and a record of the conversation.

But look at who runs the agents. In the Clubtails campaign, the workflow bypassed the demand-side platform, and the sell-side platform ran the agents. The intermediary didn’t disappear. It changed shape. Whether you can see what it takes depends on what each implementation reports, not on the protocol’s existence.

And the third problem sits outside any buying protocol. In a Vogue sponsored article we examined at the network level, the same reader was monetized three ways at once: the sponsorship, a live programmatic auction on the page and the resale of audience data, as described in the sponsored article is the top of the funnel, not the funnel. A cleaner way to negotiate a deal changes none of that. Transparency still comes from reading what a page and a supply path actually do.

Where agentic buying fits first

The early examples share a pattern. They are premium, negotiated inventory: connected TV, live sports, linear television. That is the part of media buying still run on emails, requests for proposal and insertion orders, where automating the back-and-forth saves real time. It is also where programmatic automation has reached least.

For most mid-size advertisers, those channels sit higher in the funnel than their performance spend. If CTV is part of your plan, the case for it rests on what it does downstream, the argument in why CTV is search marketing’s missing link, not on how the buy was negotiated.

What to ask before you buy through an agent

You don’t need to adopt AdCP to benefit from what it is pushing toward. You need to ask the questions it makes easier to answer.

Ask who operates each agent in the transaction and how each is paid. An agent run by a seller is working for the seller. Ask for a log of what was proposed, changed and approved, since human approval is built into how these deals are described. Ask which fees are visible in reporting and which are folded into the price. And ask how delivery is verified independently of the parties who sold it.

Then judge the result the way you would any other media: on what it caused, not on how modern the pipes were. A cheaper, faster path to the wrong audience is still waste, and the other sources of ad cost don’t go away because agents negotiated the deal, as covered in how to lower ad costs. Keeping buying decisions tied to measured outcomes, whatever the transaction layer, is the work of performance marketing measured on outcomes rather than on the buying method.

AdCP didn’t open the black box. It gave the industry a new, faster way to reach into it. What you see inside still depends on asking.

Frequently Asked Questions

What is AdCP?

AdCP, the Ad Context Protocol, is an open standard that lets AI agents representing buyers and sellers plan, negotiate and execute advertising deals in a common format. It launched in October 2025, is stewarded by AgenticAdvertising.org, runs on the Model Context Protocol and related agent standards, and covers discovery, planning, creative, audiences, governance, reporting and serve-time activation of pre-negotiated deals.

Is AdCP replacing programmatic advertising?

Not in practice. Early AdCP and agentic deals have focused on premium, negotiated inventory such as connected TV and live sports, while real-time bidding still handles most open programmatic trading. IAB Tech Lab is also extending existing standards such as OpenRTB for agents rather than replacing them. For now, agentic buying is an added lane alongside programmatic, not a substitute for it.

How does agentic buying compare with programmatic and traditional ad buying?

Traditional direct buying relies on people negotiating through proposals and insertion orders. Programmatic automates the purchase of individual impressions through auctions. Agentic buying automates the negotiation itself, with buyer and seller agents exchanging briefs and proposals, usually with human approval before execution. Many advertisers will use all three, choosing by channel: agents for premium negotiated media, programmatic for scale, direct deals for key partners.

Does AdCP make ad fees transparent?

Not automatically. AdCP standardizes how agents communicate and can create a record of proposals and approvals, but what fees and data are visible depends on what each implementation reports and who operates each agent. Advertisers should still ask who runs the agents, how they are paid, which fees appear in reporting and how delivery is independently verified. Those answers will vary by platform and deal.