There is a particular sound the ad industry makes when a standards body publishes something. It is the sound of four thousand people forwarding a PDF to each other with the words "interesting, thoughts?" and then never opening it again.

On July 8, the IAB published its Redefining Media Types standard, RMT if you like acronyms, and you people love acronyms. It is a taxonomy. It sorts video inventory into three buckets based on how a human being actually experiences the ad: lean-back big screen, personal screen, and public ambient display. Connected TV, mobile, digital out-of-home. Over the top of that it layers operational attributes for data signals, content context, device type, and addressability. A working group started meeting in January. It had measurement people, streaming product leads, holding company chief investment officers, and Media Rating Council members in the room. The public comment window closes August 8.

So far, so committee.

Here is the part that made me put down my coffee. The draft draws a hard definitional wall between CTV and DOOH, and it does it by anchoring CTV to the physical device. A television screen is CTV. Netflix on your iPad is not CTV. Netflix on your laptop is not CTV. The public screen at the airport gate is a third thing entirely, in its own bucket, with its own name, priced accordingly.

Nobody on the buy side talks like this.

I want to be careful, because I went looking for a named agency trader or a Trade Desk executive publicly telling the IAB to pound sand over this, and I did not find one in the trade press. What I found instead is more damning in its quiet way: the industry's own reference documents cannot agree with each other.

Nielsen is the outlier on paper. In practice, everybody is Nielsen. Ask a media planner what CTV is and they will say Netflix, Hulu, Peacock, Roku Channel. They will name services, not hardware. The plan is organized by where the audience went, not by what glass the pixels landed on.

The IAB is asking a market that thinks in content to start thinking in cabinetry.

And then I watched a transcript from earlier this year, and the story stopped being about the buy side at all.

Jamie Finstein is the IAB VP of the Media Center. She runs it, whatever it it. She has argued publicly for years that the industry should reserve CTV for the device and OTT for the content, which is why RMT is less a fresh invention than a long-standing position finally getting written down.

Earlier this year, on camera, only months before the IAB published the new standards, an interviewer asked her the only question that matters in her entire portfolio:

"Is CTV just becoming TV at this point?"

Here is what she said, or what I heard her say.

"Well, I think it's really what the marketer's objective is."

So, according to her, CTV is, whatever you think it is — and not whatever the IAB defines. Well now.

Then she talked about mass reach, and households, and leveraging first-party data from a retailer, and how linear and streaming serve different purposes depending on what a brand wants.

She was handed the definitional question on a plate and answered in terms of buyer objectives. Not screens. Not hardware. Not devices.

And it gets worse when you count her usage across the rest of the conversation. Her thesis in that interview is that the story of 2026 is not any single trend but convergence, everything collapsing into everything else. So:

"you don't really talk about commerce anymore without thinking about CTV." Commerce is a market, not a device.

"creator now is in the CTV space. Creators have their own programming." Programming is content. A creator's video does not become a different media type because someone picked up a phone, but under RMT that is precisely what happens: big screen is one bucket, phone is another, same creator, same ad, different category, different price.

Live is "playing an increasingly important role in the streaming space" and propelling CTV's growth. Live is content. Streaming is a distribution model. Neither is a piece of glass.

And CTV can now be measured through closed-loop attribution. Closed-loop attribution runs on identity, and identity attaches to a household or a person, never to a cabinet.

Four uses. About three minutes. Not one of them device-bound.

I am not calling her a hypocrite. That’s just too easy. I am saying something worse for the standard she is shipping. I went looking for evidence that the buy side rejects this vocabulary and could not find a documented objection.

This is better than an objection. An objection is one company's commercial position and gets discounted as such. This is the author of the framework, in her element, unprompted, with every professional incentive to model the new vocabulary, reaching four times for the usage her own framework forbids.

Because it is the usage that communicates.

Language is not a policy choice.

It is a behavior, and behavior under no pressure is the only honest sample you get. If the vocabulary cannot survive contact with its own author's mouth in a friendly interview, it will not survive contact with a planner at 4:40 on a Thursday trying to get a deck out the door.

She has a real answer, and you should have it from me rather than from a statement next week.

Look at the axis she actually used to separate linear from CTV: mass reach versus reaching a specific person with a retailer's first-party data. That is addressability, and addressability is already in RMT as an operational attribute layered on top of the media type buckets.

So there is a coherent architecture here. Media type is the container. Addressability, data signals, and content context are the attributes. Under that reading the device split never claimed to be the meaningful distinction. It claimed to be the stable one, the thing that holds still while everything else converges around it. Categories matter more when a market blurs, not less. She said fragmentation is growing. A machine parsing a bid request needs one field that does not move.

That is defensible design. It is also exactly the argument she should have made when asked whether CTV is becoming TV, and did not.

But it leaves the contradiction standing. You cannot describe a market as converging and then build your taxonomy on the premise that its categories separate cleanly by hardware. One of those two statements is load-bearing. The other is a press release.

The stated reason for all of this, right now, is AI agents. If humans cannot agree on a shared language, we cannot expect the agents transacting on our behalf to speak one. You cannot train models on a mess. You cannot have a buyer's agent and a seller's agent negotiate over a word that means two different things depending on which glossary each one swallowed.

That is a real argument, which is why I am about to be rude about the evidence.

The proof point Finstein reached for on the Digiday Podcast, hosted by Kimeko McCoy, was that three to five percent of transactions are already agentic. The source? A panelist. At a conference. In Denver. Unnamed panelist, unnamed conference. I went looking in searchable coverage and could not locate it independently.

That is a hallway rumor with a decimal point. It gets born in a Marriott ballroom, gets repeated on a podcast, and six weeks later shows up in a deck with a footnote pointing at the podcast.

The irony is she never needed it. NBCUniversal and FreeWheel already completed what has been described as the first AI-agent-led programmatic guaranteed deal, live sports and streaming inventory, buyer agent and seller agent settling terms with no human in the execution loop. The plumbing is being poured now: the Advertising Context Protocol, built on Anthropic's Model Context Protocol, plus the Unified Context Protocol and the Agentic RTB Framework. Programmatic took a decade to standardize. This is happening in quarters. The thesis stands on the record. Reaching for the ballroom statistic tells you how badly the IAB feels it needs to be believed.

Yesterday we published The Ledger Wars, six months of reporting on who ends up owning advertising's system of record once agents are moving the money. Mediaocean, The Trade Desk, the verification duopoly, and the finding that the authoritative record of delivery was settled years ago in the 4A's and IAB Standard Terms and Conditions and this industry barely discusses it.

The closing instruction was: watch the boring layer, watch who controls the schema.

Here is the schema, arriving three weeks later, wearing a lanyard.

And notice which IAB. The Ledger Wars was largely a story about IAB Tech Lab and Tony Katsur. RMT comes out of IAB proper, the Media Center, Finstein's shop. Two organizations, one brand, working opposite ends of the same corridor: one building the audit and agent-registry layer, the other defining the categories that layer will eventually be auditing. Finstein points at Tech Lab's work herself in that same interview, citing OMSDK and verification partners on signal quality, so the awareness is there. The map is not.

Katsur, asked in that piece who wins the agentic era, would not name a company. He said the winners will be whoever sits "closest to trusted operational data."

A taxonomy is where operational data gets its meaning. That is not a coincidence. That is the same fight, one floor down.

Here is my actual problem with the IAB, and it is not the one people assume.

I do not think the IAB is bad at this. The working group is serious, the attribute layer is thoughtful, and the CTV/DOOH firewall solves a real fraud problem buyers have been eating for years. The document is better than most.

My problem is that the agencies are not listening. Not because the work is wrong. Because the agencies listen to the ANA, and the ANA is where their clients sit. When a CMO forwards something to their agency, the agency reads it. When a supply-side trade body publishes a taxonomy, the agency files it.

That is not a moral failure. That is an org chart.

And it is compounded by the mechanism, which is the part I keep returning to. The IAB changes what words mean by announcement. Not by adoption, enforcement, certification, or consequence. A group meets, a document ships, a release goes out, and the industry is informed that a word it has used for a decade now means something else. Ask what happens to a seller who labels inventory the old way on August 9 and the honest answer is nothing. No fine, no delisting, no bid rejection. Finstein has effectively conceded this, telling AdExchanger that without adoption, nothing changes.

That is candor, and I respect it. It is also a confession that the standard is a suggestion with letterhead.

So the pattern repeats: the IAB does the work, publishes the standard, and the standard becomes a thing you cite in an RFP dispute rather than a thing that changes what gets bought. A reference document. A rhetorical weapon. Useful. Not transformative.

Except.

Except this time I do not think the pattern holds, and not for the reasons the IAB is arguing. It has nothing to do with whether traders relearn a word for Netflix on an iPad. It has to do with what a taxonomy becomes once machines start reading it, and what happens when the boring layer quietly turns into the thing everyone downstream is forced to obey.

Rio Longacre gave us the sentence for it yesterday, about execution platforms: "An execution agent should not be allowed to grade its own homework." His warning was that if the platform running the buy becomes the default source of truth, we rebuild every conflict already in this ecosystem, at machine speed.

The same warning applies one layer up. If the party defining the categories has a commercial stake in which categories exist, you do not need an execution agent for the conflict to arrive. It is already baked into the vocabulary.

That is the real fight. Not CTV versus DOOH. Not device versus content.

Who owns the system of record that AI consults, trusts, and acts on.

ADOTAT has no financial relationship with any company named in this piece: not the IAB, IAB Tech Lab, Digiday, AdExchanger, The Trade Desk, Netflix, Nielsen, Amazon, Google, Mediaocean, VideoAmp, iSpot, NBCUniversal, FreeWheel, Anthropic, or any agency holding company. No subject received advance review of this article, no sponsor had any input into its reporting or framing, and this is not sponsored content. This publication is funded by subscriptions and reader support, which is the entire reason it can print things that make its own industry uncomfortable.

The argument about CTV is the part everyone will tweet. The part behind ADOTAT+ is where we trace the wiring underneath it. We follow the standard from the moment it leaves an IAB working group to the point where it becomes a machine-readable field, a contractual definition, a measurement dispute, and eventually a line item a CFO signs off on. Along the way we identify who gains, who loses, which companies are quietly positioned to become the judges of AI-driven advertising, and why this fight has far less to do with television than with who gets to define reality before the software starts spending.

We also name the edge cases nobody wants to touch, explain why hotel TVs, airport screens, bar televisions, and FAST channels expose cracks in the standard, examine whether agencies will actually adopt it, and make a prediction about what changes after the August 8 comment period. Most importantly, we connect this story to The Ledger Wars and show why the taxonomy, the protocols, and the system of record are not separate battles at all. They're three fronts in the same campaign. Almost nobody in advertising has drawn the whole map. We did.

Upgrade here