The most useful thing anyone can do for the agentic advertising category right now is publish real numbers, because almost everything said about it is projection. This piece establishes what an agentic advertising index should measure, reads the marketplace signals that are already public, and sets out the metrics the category needs to be tracked on over time. It is the first entry in an ongoing index, and its purpose is to move the conversation from forecasts about what agents will do to measurement of what they are actually doing.
TL;DR. The agentic advertising category is measured almost entirely in forecasts and adoption intent, not in outcomes. An index worth the name should track four things: adoption depth (not just who is planning agentic buying but who is running it in meaningful production), execution reality (live agent-to-agent buys, the standards they run on), efficiency effects (what agentic buying does to working media and supply-path waste), and reconciliation integrity (whether agents transacting actually agree on what they agreed). The public data today shows broad adoption intent, a handful of real live buys, and maturing standards, but almost no outcome data, which is exactly the gap an index exists to fill. What gets measured is what the category will be able to argue about with evidence rather than assertion.
Why does the category need an index?
Because the ratio of claims to data is dangerously high. Agentic advertising is discussed almost entirely in the future tense: what agents will do, how much they will change, where the market will be in a few years. The forecasts are useful, but they are forecasts, and a category that runs on projection is a category where the loudest vendor wins the narrative regardless of what is true. An index shifts the ground from prediction to measurement, and measurement is what lets buyers, and the trade press, tell the real from the asserted. It is worth noting that when a trade reporter recently put six direct questions to the architect of one leading protocol, the honest answer was that the binding transaction record exists and works but is not yet active in production at scale, precisely the kind of measured statement the category needs more of.
The need is sharper in agentic advertising than in most emerging categories for two reasons. First, the technology acts autonomously, so the gap between "the platform can do this" and "the platform reliably does this well" is wide and consequential, and only outcome data closes it. Second, the category is being defined in public by parties with a commercial interest in the definition, which means neutral, measured numbers are unusually valuable precisely because so much of the surrounding discourse is not neutral. The same reporting found the one operator willing to name a figure was running roughly two to three thousand dollars a day in live agent-bought volume while others issued press releases off far smaller tests, a gap that only real numbers expose. An index is a small act of discipline in a conversation that badly needs some.
What should an agentic advertising index measure?
Four dimensions, each chosen because it distinguishes what is real from what is claimed.
The first is adoption depth. The commonly cited adoption figures measure intent: how many buyers are live, testing, or planning agentic campaigns. That is a real signal, and it is substantial, industry research puts a large majority of digital video buyers in one of those three states. But intent is not depth. The more revealing metric is how many of those deployments are running safely in meaningful production versus how many are pilots or autonomous-looking systems that are not yet trusted with real decisions. Enterprise research suggests the gap between broad adoption and safe production is large, and an index should track that gap over time, because closing it is the real story of the category maturing.
The second is execution reality. This measures what is actually happening in live buying: the number and scale of genuine agent-to-agent media buys, which standards they run on, and whether agents are negotiating deals or merely automating bids. The public markers here are concrete but few, the first cross-platform agent-to-agent buy, holding companies executing live agent-to-agent buys for clients, and an index should track their accumulation, because the count and scale of real buys is the ground truth beneath all the adoption intent.
The third is efficiency effects. This measures what agentic buying actually does to the outcomes the category claims to improve: working media as a share of spend, supply-path efficiency, waste. The industry already has a baseline to measure against, since the Association of National Advertisers has tracked programmatic working-media efficiency and waste for years, and its benchmarking has shown working-media efficiency reaching new highs even as total waste remained substantial. An agentic index should track whether agentic buying moves those numbers, because "agents make buying more efficient" is a testable claim and the test is whether the efficiency metrics improve where agents are deployed.
The fourth is reconciliation integrity, and it is the one no other index measures. This tracks whether agents transacting actually end up agreeing on what they agreed: whether the buyer's record and the seller's record of a deal match, and stay matched, across a campaign. It matters because it is the category's central structural risk and because it is invisible in every other metric, a buy can look successful on adoption, execution, and even efficiency while the two sides quietly hold divergent records of it. This is exactly the failure mode IAB Tech Lab's standards leads describe when they insist an agentic transaction needs a deterministic, auditable record rather than a model asserting a buy happened.
What does the public data show right now?
It shows a category real in adoption and execution, maturing in standards, and almost empty in outcome data, which is precisely the shape that makes an index necessary.
On adoption, the intent is broad and the depth is thin. A large majority of digital video buyers report being live with, testing, or planning agentic campaigns, and holding companies have committed serious resources, including a multi-billion-dollar acquisition explicitly positioned to fuel agentic buying. But the enterprise evidence indicates only a minority of agentic AI deployments are running safely in meaningful production, so the adoption-depth metric, the one that matters, sits well below the adoption-intent metric that gets quoted.
On execution, the reality is genuine but early. Live agent-to-agent buys have happened, including a first cross-platform buy with buy-side and sell-side agents negotiating over a shared protocol, and major agencies have run live agent-to-agent buys for clients. These are real and they are documented. They are also, so far, a small number of notable events rather than a routine volume, which is exactly what an execution-reality metric should track as it grows.
On standards, the maturation is rapid and measurable. The IAB Tech Lab's agentic specifications have moved from announcement to shipping versions with working reference implementations, extending established standards rather than replacing them, and a competing framework is developing in parallel. The pace of standards development is itself a category-health signal, and it is currently strong.
On efficiency and reconciliation, the data is close to absent, and that absence is the point. There is little public outcome data on what agentic buying does to working-media efficiency, and almost none on reconciliation integrity. The only substantial signal on the latter comes from simulation rather than live measurement: published research modelling agentic transactions against current specifications found that two agents which had just agreed a deal recorded its terms differently in the large majority of cases. That research models the specification architecture rather than any production system, and no platform has published live reconciliation data of its own, which is exactly the kind of gap the index is meant to expose and, over time, fill.
What should buyers do with an index like this?
Use it to replace vendor narrative with measured reality, and to ask vendors for the numbers the index tracks. When a platform claims broad adoption, the index-informed question is how much of that adoption is in safe production versus intent. When a platform claims efficiency gains, the question is what happened to working-media share, measured against the established baseline. When a platform claims its agents transact reliably, the question is what its reconciliation data shows, and if it has none, that absence is itself information.
The broader value of an index is that it makes the category accountable to evidence over time. A single snapshot is a starting point; the real worth accumulates as the same metrics are tracked across quarters and the trajectory becomes visible, whether adoption depth is closing the gap on adoption intent, whether live buys are becoming routine, whether efficiency is actually improving, whether reconciliation integrity holds as volume grows. Those trajectories are what will tell buyers whether agentic advertising is maturing into a reliable way to buy media or stalling as an over-promised experiment.
This is the first entry in that record. Its numbers are necessarily partial, because the category's own outcome data is partial, and saying so plainly is part of the discipline. The purpose is not to declare where agentic advertising has arrived, but to start measuring it honestly, so that the next entries can show movement against a real baseline rather than adding another forecast to a category that already has too many.
This article references industry adoption research, public reporting on live agent-to-agent media buys and holding-company investment, reporting on what agentic buying can currently prove, the IAB Tech Lab's agentic specifications, the Association of National Advertisers' programmatic transparency benchmarking, and published simulation research into agentic deal reconciliation conducted by Alkimi Marketplace. The simulation models the current specification architecture and is not an assessment of any specific production platform.