The tidy story the industry has been telling itself about agentic advertising standards took a direct hit on 19 August, when Shailley Singh, COO and EVP of product at the IAB Tech Lab, published a pointed clarification of what AAMP — the Agentic Advertising Management Protocols — actually covers. The comfortable framing, repeated across vendor blogs and conference panels, holds that AdCP (the Ad Context Protocol, governed by AgenticAdvertising.org) handles the campaign-planning layer while AAMP handles impression-time execution: complementary layers, no conflict, implement both. Singh’s post says that map is drawn wrong. “AAMP is not an ‘impression-layer protocol,’” he writes — it is an umbrella framework spanning campaign planning, media and audience discovery, transaction management and impression-time execution, with linear TV buying included for good measure.
The overlap he documents is not marginal. AAMP and AdCP overlap on 13 business functions by the Tech Lab’s own accounting, from natural-language campaign briefs to deal creation. The list covers buyer and seller agents, inventory and seller discovery, media planning, pricing and negotiation, human approval gates before spend, audience activation, agent identity and registry functions, campaign lifecycle reporting, and serve-time decisioning. Singh’s structural point: AAMP comprises three pillars — Agentic Foundations, which includes the newly finalised Agentic Real-Time Framework (ARTF); Agentic Protocols and Standards, carrying the buyer and seller agent SDKs; and the Agent Registry for trust and transparency. The recurring analytical error he calls out is treating ARTF as the whole of AAMP, then slotting that reduced version neatly underneath AdCP.
He also dismantles the softest argument for complementarity: that both stacks speak MCP and A2A, so they must fit together. “MCP is a communication substrate; it does not reconcile their advertising semantics,” Singh writes. Two protocols sharing a transport is like two companies sharing a phone network — it says nothing about whether they agree on what a deal, an audience or a price actually is.
01A standards war, now officially acknowledged
Readers of our earlier coverage will recognise the trajectory. We covered AAMP 2.3 in our 2 August issue as the moment agentic buying got real guardrails — verified inventory, provable pricing, human approval gates. AdCP, meanwhile, launched on 15 October 2025 with six founding members and 23 additional participants, notably without Google, Amazon, The Trade Desk or Microsoft. For most of this year the two bodies maintained a public posture of peaceful coexistence, and some players hedged accordingly — VIOOH, whose seller agent automated over a hundred curated DOOH deal requests in the first half, planned to register under both rather than pick one.
Singh’s post ends the polite fiction. When the standards body itself writes a rebuttal titled around why the comparisons miss the mark, the subtext is plain: these are two substantially overlapping visions for how software will buy advertising from software, and the industry will eventually converge on one semantic layer per function, not two. That is not a criticism of either body — early standards competition is normal, and it usually produces better specifications than a premature monopoly would. But it changes the advice. “Implement both, they’re different layers” was cheap counsel when the layers were believed distinct. Thirteen overlapping functions means dual implementation is duplicate integration work, done twice, against semantics that may diverge further.
02What publishers should implement, and what can wait
The practical question is where a publisher’s effort survives either outcome. The answer is the parts that are about your house rather than either protocol’s semantics. Clean, machine-readable inventory declarations, accurate product catalogues, honest pricing structures and reliable reporting are prerequisites for any agentic buyer under any standard — that work transfers entirely. Protocol-specific seller-agent integrations are the part that may need redoing, and that argues for sequencing: adopt what your existing stack already supports, pilot narrowly where a real counterparty is transacting today, and keep the deep integration spend until the semantic layer settles.
03Why this matters for publishers
| The "implement both" advice just got more expensive | The Tech Lab's own list puts the overlap at 13 functions across the two frameworks. Dual integration therefore means maintaining two semantic mappings of your inventory, deals and pricing — and reconciling them when they disagree. |
|---|---|
| Whoever defines the semantics defines your shelf placement | How a protocol represents audiences, deal types and pricing determines how agentic buyers see and rank your inventory. A standards war over those definitions is a war over how you get bought. |
| Governance is part of the choice | AAMP sits inside the IAB Tech Lab's membership process; AdCP sits with a newer consortium that launched without several of the largest platforms. Where the big demand sources eventually land will matter more than any feature comparison. |
| Shared plumbing is not shared meaning | Singh's MCP point generalises: vendors waving protocol-compatibility claims at you may only be describing the transport. The questions that matter are who verifies agent identity, how pricing is proved, and where human approval sits. |
04What publishers should do
05The bottom line
Singh’s post is the moment the agentic standards conversation stopped pretending to be a story about complementary layers and became what it structurally is: two overlapping frameworks competing to define the semantics of machine-to-machine media buying. For publishers, the update is clarifying rather than alarming. The war is over the protocol layer, not over whether agentic buying happens — and the preparation that matters most, making your inventory legible and your pricing provable, is identical under every outcome. Do that work now, pilot where real spend already flows, and let the two standards bodies spend their own money settling whose semantics win.