As AI agents take over more media planning and buying, publishers and streaming operators need a standard way to show their inventory to buyers that aren’t human. AAMP 3.0 is IAB Tech Lab’s attempt to provide one.
Publishers have spent years polishing media kits, rate cards, and RFP responses for human readers. The next reader may be software. Two-thirds of US ad buyers told IAB, in its 2026 Outlook Study, that they plan to focus more on agentic AI for media buying and campaign management this year, and half of US marketing agencies already use agentic AI for marketing execution, according to Forrester and the 4A’s.
The supply side is moving more slowly. A Digiday and Optable survey of 180 industry professionals found that 89% of advertisers are using or building agentic AI, while only 29% of publishers are actively deploying it. Publishers listed technical complexity (68%) and unclear standards (41%) among their biggest barriers. In other words, marketers are getting ready to send AI agents to media owners, and many media owners aren’t ready to talk to them.
That is the gap AAMP 3.0 is meant to close. IAB Tech Lab released it on September 22, 2026, less than two months after version 2.3. The main new piece is OpenProposal, a format that lets publishers describe their ad products and answer buyer briefs in a way AI agents can read. But there are reasons to be careful: OpenProposal is still a draft, not an approved standard, and the industry hasn’t decided which agent protocols will win. Below, we look at what AAMP 3.0 actually changes for publishers and streaming operators, and where the hype may run ahead of reality.
What AAMP 3.0 Actually Adds
So far, most of AAMP’s work has been about execution: making the standards the industry already uses for deals, auctions, and delivery work with AI agents. AAMP 3.0 moves one step earlier, to the point where an advertiser’s brief meets a publisher’s offer. Whether it’s a standard package or a custom response, an OpenProposal can spell out what’s included, who it reaches, the pricing and terms, the guarantees, and how it will be delivered. Once both sides agree, each line item moves into the standard it maps to: OpenDirect 2.1, Deals API, or OpenRTB 2.6. Nothing already in production has to be replaced.
The case for it is mostly about scale. A human media team can only review proposals from a handful of publishers per campaign, so most of the market never sees the brief. AI agents could change that, but only if publishers describe their inventory in the same structured way. That’s why the format is built the way it is: out of 71 fields, only three need a back-and-forth between the advertiser and the publisher.
What It Means for Publishers and Streaming Operators
The first change is in how inventory gets described. What’s new, according to PMG’s head of CTV strategy in the launch announcement, is that the reasoning travels with the proposal: a publisher can explain why a package fits a brief, including the audience, the formats, and how the price was set.
For streaming operators, this context may matter even more. The value of a CTV impression depends on the content around it, the ad load, the device mix, and whether the event is live, and most of that doesn’t fit into a rate card. A format that can carry those details could help premium streaming inventory stand out from look-alike supply. How well OpenProposal handles these streaming details is something the comment period should show.
The practical work is less glamorous than building an AI agent. As per the draft specification, OpenProposal is meant to run alongside a publisher’s order management system and ad server, so a proposal is only as trustworthy as the availability data behind it. The same survey backs this up: only 5% of respondents planned to prioritize building or enhancing AI agents over the next year, while 49% were focused on improving first-party data and signals so outside agents can find them.
Adoption Perspectives: Where the Hype Meets Reality
OpenProposal solves a real problem, but it’s still early. IAB Tech Lab extended the public comment period due to strong demand, and it closed on November 10. The specification remains a draft, though: its repository warns against building production systems on it for now, and the machine-readable schemas are expected to arrive with the first revision. It also has to compete with AdCP, and Adform’s CTO argues that neither protocol runs at scale yet. Trust is limited too: in the IAB survey mentioned earlier, 75% of advertisers aware of agentic AI use it or are likely to use it for inventory discovery and evaluation, but only 45% for direct IO negotiations. Agents may be reading proposals long before anyone trusts them to close a deal.
Still, groundwork is under way. IAB Tech Lab says it will build OpenProposal into its Buyer and Seller Agent SDKs. IAB Australia, meanwhile, has been testing the draft in its own lab and expects the first seller agents to run in a sandbox in mid-to-late November, according to its own report. We expect the first six months after the standard is finalized to belong to large advertisers, platforms, and the biggest publishers. Smaller publishers will likely come in through their ad tech vendors rather than build their own agents.
For publishers and streaming operators, the practical move is to get the basics ready before an agent asks: structured product descriptions, accurate availability, and an ad server that can take demand from several sources. That’s the idea behind the AdPlayer.Pro CTV Ad Server, which supports real-time OpenRTB auctions, multi-demand routing, and ad podding, with live analytics and debugging tools. Standards will keep changing, but clean, structured, up-to-date inventory data will stay useful whichever protocol wins.
AI Agents Are Coming to Programmatic: What AAMP 3.0 Means for Publishers