Gravity is Relative

When IAB Tech Lab CEO Anthony Katsur says "when you remodel the kitchen, you don't burn down the house," he's making a structural argument, not a metaphor. AAMP (Agentic Advertising Management Protocols) is that argument rendered in specifications, SDKs, and a deliberately unglamorous three-pillar architecture that does something AdCP cannot: it preserves everyone's existing business model.

But whose existing business model? IAB Tech Lab governs the open programmatic ecosystem, which accounts for roughly 42% of non-China digital ad spend. The other 58% sits inside walled gardens (Google, Meta, Amazon) that don't use IAB Tech Lab standards internally and have no structural need to. When we talk about AAMP as the "centre of gravity," we're talking about the centre of gravity on a moon. The planet has its own physics entirely.

That context matters, because the existential question facing both AAMP and AdCP isn't which one wins. It's whether the open ecosystem they're both trying to govern retains enough commercial gravity to justify the engineering investment. Every year the attribution gap between walled gardens and the open web persists, more budget migrates to closed-loop platforms where measurement is deterministic and AI optimisation is structurally superior.

The institutional play

Start with the obvious: AAMP is an institutional defence. If AdCP (or any other body) becomes the canonical standard for agentic advertising, IAB Tech Lab loses its central role in the ecosystem it has governed for over a decade. AAMP isn't just a technical response; it's a jurisdictional claim. The ExchangeWire article published today makes it explicit: "all and any agents emerging from the various projects, for example AdCP, should be registered with the IAB Tech Lab Agent Registry." That's not interoperability. That's annexation by invitation.

The Agent Registry is the strategic centrepiece, not ARTF. Whoever controls the registry controls discoverability and trust. It's the ads.txt playbook applied to agents: low friction to adopt, high lock-in once it becomes the canonical lookup. Amazon has already registered. PubMatic and Equativ too. Ten entries by mid-March, with a three-tier deployment classification (Remote, Local, Private) that gives the registry genuine operational utility beyond a simple directory.

Why TTD supports AAMP and not AdCP

The most telling data point in the entire agentic standards debate is a single quote from Arpad Miklos, Staff Software Engineer at The Trade Desk, praising ARTF as "a sensible and empowering approach to opening up the augmentation space." TTD is actively implementing it.

TTD's hostility to AdCP and embrace of ARTF are two sides of the same coin. AdCP reduces the buyer to a natural-language brief and positions all intelligence on the sell side. ARTF does the opposite: it containerises third-party agents within the existing bidstream infrastructure, preserving the DSP's central position as the decisioning node. The buyer's intelligence (identity resolution, SPO, frequency management) remains intact. The seller gets faster, richer auctions. Nobody's role is eliminated.

This is why "evolution vs revolution" isn't just a philosophical distinction. It's a question of who keeps their margins.

What AAMP actually solves

Strip away the territorial politics and AAMP has genuine technical merit, though less novelty than its proponents claim. ARTF's containerisation model addresses a real infrastructure problem: the latency and data-leakage costs of external HTTP calls in the bidstream. Moving identity resolution, fraud detection, and audience enrichment into co-located containers within the host platform cuts round-trip times from 400-800ms to roughly 100ms. Chalice AI has been testing this inside Index Exchange with real results.

But the concept is not new. Beeswax (acquired by Comcast/FreeWheel in 2022) offered "bidder-as-a-service" years ago, letting buyers deploy custom algorithms on shared infrastructure co-located with exchanges. Header bidding already containerises multiple demand sources. Index Exchange and Chalice were running co-located models before ARTF existed; the spec was partly written to standardise what they were already doing. What ARTF adds is the portable, standardised interface (OpenRTB Patch, gRPC/protobuf, container manifests) that makes this "package once, deploy to any compliant host." The innovation is in the interoperability specification, not in containerisation itself.

The "agentified" standards approach (Agentic Direct, Agentic Bid, Agentic Audiences) is pragmatic rather than exciting. It extends OpenRTB, OpenDirect, and AdCOM with MCP and A2A support rather than defining new task schemas. The advantage is that every DSP, SSP, and ad server already speaks these object models. The disadvantage is that you inherit all of their accumulated complexity and compromises.

The bias nobody discusses

AAMP is presented as neutral: a non-profit standards body serving the whole ecosystem. But IAB Tech Lab's membership and funding come from the companies whose existing infrastructure AAMP is designed to preserve. The standard optimises for backwards compatibility because the membership demands it. The companies paying the bills are the ones with the most to lose from disruption.

This isn't corruption; it's structural incentive alignment. But it does mean AAMP is unlikely to produce standards that fundamentally challenge the DSP/SSP/exchange model, even if the technology makes such a challenge possible. AdCP, for all its sell-side bias, at least has the ambition to imagine a different architecture. AAMP's ambition is to make the current architecture faster and more autonomous. Whether that's sufficient depends on whether you think the current architecture's problems are operational (fixable with better plumbing) or structural (requiring new plumbing entirely).

Where this lands

Both AAMP and AdCP have institutional biases baked into their design. AdCP is a sell-side protocol that routes around buy-side intelligence. AAMP is an incumbency-preservation protocol that routes around structural disruption. Neither is neutral. Both are useful.

But both are also fighting over a shrinking pie. The walled gardens that control 58% of non-China digital ad spend have their own agentic infrastructure (Google's AI Max, Meta's Advantage+, Amazon's Rufus) operating in closed loops that structurally outperform open-web alternatives on attribution and optimisation. The open ecosystem doesn't need to defeat the walled gardens, but it does need to close the measurement gap fast enough to retain viable investment. Whether AAMP or AdCP (or both, or neither) can deliver that is the question that matters more than which protocol wins the standards war.

For an AI platform evaluating which standard to engage with, the Agent Registry is the component worth watching most closely. If it reaches critical mass, it becomes the canonical trust layer for agentic advertising regardless of which protocol agents use to transact. That's a more powerful position than any individual protocol specification. Just remember that the registry governs a moon, not the solar system.