On 13 November, IAB Tech Lab opened public comment on its Agentic RTB Framework v1.0. The release came weeks after a rival consortium published AdCP, and the industry now has two incompatible visions of how machine agents will buy media in 2026.
CEO Anthony Katsur framed ARTF as evolution rather than rupture. It extends OpenRTB, AdCOM and VAST, colocates bidding logic in containers, and promises latency cuts from 600 to 800 milliseconds down to roughly 100. AdCP takes the opposite path. Launched in October by PubMatic, Scope3, Swivel, Yahoo, Optable and Triton, it routes transactions through seller agents that bypass DSPs entirely. The omissions matter. Trade Desk, DV360 and Amazon DSP are absent from the AdCP roster, and Swivel is targeting twenty publisher seller-agent integrations by year-end.
This is not a technical disagreement. It is a fight over who sits in the middle of the next decade of advertising.
Two protocols, two power structures
Read ARTF and AdCP side by side and the politics become obvious. ARTF preserves the demand-side stack. It accelerates the existing pipes, keeps DSPs central, and gives buyers more speed without forcing them to rewire procurement. AdCP rewrites the topology. Seller agents negotiate directly with buyer agents, compressing the supply chain and shifting margin toward publishers and the SSPs that orchestrate them.
Neither side will concede in 2026. ARTF has the weight of OpenRTB's installed base and the Tech Lab's standards machinery. AdCP has the speed of a coalition that does not need consensus from the entire ecosystem to ship. Both will harden their codebases over the next two quarters. Both will court the holdouts. Buyers who pick a side now are betting on a winner in a market that has not yet defined what winning looks like.
The cost of picking wrong
The instinct in adtech is to align early. Pick the standard, integrate, claim the leadership narrative. That instinct will be expensive this cycle.
If you wire your buying stack exclusively into ARTF, you inherit the assumption that DSPs remain the locus of decisioning. If AdCP's seller-agent model gains traction with the publishers you actually need, you will find yourself routing around your own infrastructure within eighteen months. The mirror risk applies. Commit to AdCP and you cede speed advantages to competitors operating on the faster, latency-optimised ARTF path, while betting that the three largest DSPs in the market never join the consortium they were excluded from.
Neither bet is defensible on the evidence available. Public comment on ARTF is still open. AdCP has not yet published a major DSP signatory. The honest answer is that we do not know which protocol will dominate, and anyone who claims otherwise is selling something.
Architect for both, commit to neither
The defensible posture for 2026 is protocol-agnostic orchestration. Treat ARTF and AdCP as interchangeable transport layers and build the decisioning, measurement and creative logic above them. This is the design principle behind Xanite. Our orchestration layer abstracts the protocol so that a campaign brief, an audience definition and a creative variant route to whichever agent surface delivers the outcome, regardless of whether the handshake speaks ARTF or AdCP.
The architectural discipline is straightforward. Keep your audience graph, attribution model and creative management system independent of the bidding protocol. Negotiate seller-agent pilots with the AdCP publishers that matter to your category, while maintaining ARTF-compliant DSP relationships for scale. Instrument both paths against the same outcome metric. Let performance, not allegiance, decide where spend lands quarter by quarter.
What to do before year-end
Three moves are worth making now. First, audit every DSP and SSP contract for protocol exclusivity clauses, because the lock-in language being drafted today will constrain you in eighteen months. Second, run a parallel pilot. Allocate a small budget to an AdCP seller-agent integration and a matched budget to an ARTF-ready DSP, then compare latency, win rates and outcome metrics on the same audience. Third, name an internal owner for agentic protocol strategy before the consultancies do it for you.
The teams that win 2026 will not be the ones who guessed the protocol correctly. They will be the ones who built the orchestration layer that made the guess irrelevant.