Agentic Buying Standards Fight: AdCP vs. IAB Tech Lab Examined

Full analysis

Two AI groups are fighting over who writes the rulebook for AI agents that buy ads on their own. AdCP, backed by the labs and DSP-side engineers, versus the IAB Tech Lab, the industry's incumbent standards body. Pesach Lattin put the same six questions to both camps and, by his account, got real answers from each, which tells you the fight is live. For an ad-tech operator, the question is whether to build to a standard now, wait, or hedge, and what it costs you if you guess wrong.

Reversibility: Type 2 for the wait-and-see part, Type 1 once you commit engineering to a spec. Picking a protocol and wiring it into your bid path is expensive to unwind. Waiting is cheap until it isn't.

What's actually being decided: Not "which standard is better." It's whether agentic buying clears production at scale soon enough that the standards choice matters this budget year, and if so, whether you build to one, both, or neither.

Forcing function: None hard yet. No mandated compliance date, no publisher ultimatum. The urgency is manufactured by the news cycle, not by a real deadline.

The Skeptic. In plain terms: the industry has fought standards wars before, and they usually end in two half-built systems and a lot of panels. Most "agentic" buying today is rules-based bidding with a language model stapled to the brief. Autonomous agents placing real media without a human on the hook? Nowhere near mainstream. The binding constraint isn't the protocol. It's that brand safety, legal sign-off, and IO execution still need a person who gets fired if it goes wrong. Past IAB Tech Lab initiatives took two years or more to reach real SSP adoption. So the "pick a side now" framing is hype running ahead of the plumbing.

The Market Analyst. Public ad-tech shares aren't treating this as a coin-flip event yet, and they're right not to. But there's a quiet risk here for the independent exchanges, Magnite and PubMatic, if agent-driven volume consolidates through a protocol that favors the walled gardens. The middle of the pipe gets a slow squeeze, and it shows up in yield data long before it shows up in a headline. For a generalist: the companies that sit between buyer and publisher make money on the spread, and a new standard can widen or erase that spread. The Trade Desk is the tell. It wrote its own rules before with UID2 and OpenPath, so it gains most if AdCP wins and loses most if the IAB Tech Lab consolidates authority.

The Operator. Tuesday morning, the problem isn't philosophy, it's deal IDs. Agents don't negotiate the way human traders do. If AdCP and the IAB Tech Lab ship incompatible object models for how an agent talks to an SSP, publishers will reject the non-compliant bid requests silently, and nobody notices until yield dips. Build to both and you double the integration work with no promise either survives. Ninety days into a dual-build, engineering backlogs balloon with conflicting spec tickets, and finance gets a nasty surprise when CPMs from the agentic channel won't reconcile against legacy line items. Waiting feels safe. It's a decision with a cost, just a hidden one.

The Customer / End User. Here the customer is the buyer letting an agent spend the money, and the publisher accepting the bid. Ask the plain question: is anyone actually demanding autonomous agents with no human sign-off? The CMO who signs the media plan still owns the brand-safety blowup. Until a buyer will let an agent commit budget without a human approving the brief, the standard governs a workflow almost nobody runs at full autonomy. Publishers, for their part, care about exactly one thing: does the agent pay real CPMs and not degrade the experience. They'll adopt whichever protocol the money arrives through. They will not adopt one on principle.

Where the council splits.

The Skeptic and the Strategist disagree on the clock. The Skeptic says agentic buying is years from production scale, so the standards fight is a sideshow. The Strategist says whoever ships the reference implementation first becomes the de facto clearinghouse, so being early is the whole game. Both can't be right about timing, and the timing is the decision.

The Operator and the Customer disagree on who moves first. The Operator worries about silent bid rejections forcing an early build. The Customer says no buyer is handing an agent the checkbook yet, so the volume that would trigger those rejections isn't there. If the Customer is right, the Operator's backlog panic is premature.

The Market Analyst and the Skeptic agree the market isn't pricing this, but draw opposite lessons. The Analyst sees a real, slow squeeze on the independent exchanges worth positioning against. The Skeptic sees noise not worth a trade.

What it hinges on. Two beliefs, and everything downstream follows from them. First: does autonomous agent buying, with no human on the brief, clear real production volume inside this budget cycle? Second: will AdCP and the IAB Tech Lab produce genuinely incompatible object models, or will they converge the way most plumbing standards eventually do because the buy side hates building twice?

The council leans skeptical on urgency and pragmatic on outcome. Build a thin abstraction layer between your bid logic and whatever spec you speak, so you can switch or dual-run without ripping up the pipe. That's cheap insurance against a Type 1 mistake, and it's the one move that pays off no matter who wins.

What to verify before you spend real engineering: how much of your incoming "agentic" demand is truly human-out-of-the-loop versus a language-model wrapper on the same old rules. If it's the latter, the standards fight can wait. If a real buyer is about to let an agent commit budget against your inventory, that's your forcing function, and only then.

Prediction: Neither AdCP nor the IAB Tech Lab will have a single dominant, production-adopted agentic-buying standard by the IAB Tech Lab's next major standards update cycle; the major independent SSPs will be building dual-compliance rather than committing to one.

Confidence: Medium. Every prior IAB standards fight resolved slowly and into coexistence, not a clean winner.

Why: The last several IAB Tech Lab standards initiatives took two years or more to reach meaningful SSP adoption, and this fight started with two live, well-funded camps rather than one clear frontrunner. When the buy side faces two incompatible specs, its consistent behavior is to make vendors support both rather than bet the pipe on one, which is exactly what happened through the OpenRTB and header-bidding transitions. A clean, fast winner would require one camp to lock in the walled gardens and the big independents simultaneously, and nothing in the current split, with labs and DSP energy on one side and publisher trust on the other, points that way. The opposite outcome, a decisive standard inside a year, would break the entire track record of how ad-tech plumbing gets adopted.

Revisit by 2027-02-19: We're right if the major independent exchanges are supporting or building toward both AdCP and IAB Tech Lab agent specs, or neither has meaningful production volume. We're wrong if one standard is clearly dominant and adopted across the top independent SSPs by then.