Three weeks ago, a consortium of advertising technology companies launched the Ad Context Protocol (AdCP), an open-source standard designed to enable AI agents to communicate across advertising platforms and execute advertising tasks autonomously.

The announcement drew immediate scrutiny from industry observers questioning whether the sector needs another protocol when existing standards remain underutilised.

Then, just days ago, the IAB Tech Lab announced the User Context Protocol (UCP) — a complementary but distinct standard focused on how AI agents exchange user-related signals.

The rapid emergence of these protocols reflects advertising’s entry into what industry leaders call the “agentic era,” where AI assistants autonomously discover inventory, negotiate deals, and optimise campaigns using natural language.

For news publishers, these developments raise practical questions: Should you implement these protocols? Build for them? Plan for them? Or simply wait and see?

The answers depend on understanding what these protocols actually do, who’s behind them, and — most critically — whether there’s any real market demand to capture.

What AdCP and UCP actually are

AdCP functions as “OpenRTB for the AI era,” enabling agent-to-agent negotiation rather than impression-level auctions. Built on Anthropic’s Model Context Protocol (MCP), it consists of four core modules: the Signals Activation Protocol for audience discovery, the Media Buy Protocol for campaign execution, the Creative Protocol for asset management, and a forthcoming Curation Protocol scheduled for Q2 2026.

The technical architecture differs fundamentally from real-time bidding. Where OpenRTB handles sub-100-millisecond impression auctions, AdCP operates asynchronously — responses can take seconds or days — accommodating human-in-the-loop approvals while AI agents negotiate complex deal terms.

A publisher implements AdCP by building an MCP-compatible agent that responds to natural language briefs like “Find women interested in rock climbing in the UK” and packages relevant inventory, audience data, and pricing into standardised responses.

UCP, introduced by IAB Tech Lab earlier this month, addresses a different challenge. Rather than orchestrating media transactions, UCP standardises how agents exchange identity signals, contextual signals, and performance data. According to IAB Tech Lab CEO Anthony Katsur, “UCP is a derivative of AdCP, whereas AdCP focuses on direct media orchestration, and UCP focuses on audience and ID orchestration.”

UCP leverages embeddings — compact vector representations that efficiently encode complex signals in privacy-preserving formats. This approach aims to enable real-time signal exchange while supporting privacy requirements through integration with existing IAB standards like the Transparency and Consent Framework and Global Privacy Protocol. LiveRamp donated the protocol to IAB Tech Lab, which now governs it through an open-source commit group.

The two protocols are positioned as complementary: AdCP handles the “what” and “how” of advertising transactions, while UCP handles the “who” and “when” of audience signals. In theory, a buyer agent using AdCP to negotiate a campaign could simultaneously use UCP to exchange audience intelligence with seller agents.

The compelling promise — and concerning gaps

The value proposition for news publishers centers on disintermediation and first-party data monetisation. AdCP enables news publishers to expose inventory and audience data directly to buyer agents, potentially bypassing layers of SSPs, exchanges, and data brokers that extract fees. The protocol supports unlimited product flexibility — publishers can offer any advertising product requested in a brief, no matter how niche, without being constrained by manual packaging.

For news publishers with differentiated first-party data, these protocols could enable more sophisticated monetisation. Rather than being forced into CPM-based open exchange pricing, publishers could transact on audience segments, engagement rates, or brand-lift outcomes. The asynchronous design accommodates complex negotiations compressed RTB timelines can’t support.

However, the current reality reveals troubling gaps.

AdCP launched with six founding members (Yahoo, PubMatic, Scope3, Swivel, Triton Digital, and Optable) and 20+ launch members heavily weighted toward the supply side. Conspicuously absent: The Trade Desk, Google DV360, and Amazon DSP — the three dominant demand-side platforms. No major advertiser brands, no agency holding companies, and no walled gardens appear in the consortium.

Most revealing, according to Permutive co-founder Joe Root, zero dollars are currently transacting through AdCP. The protocol exists primarily as a conversation starter rather than actively adopted infrastructure. No AI assistant platforms have integrated AdCP despite the protocol being designed explicitly for AI agents.

This creates a chicken-and-egg problem: News publishers can build AdCP agents, but there’s no clear path for those agents to interact with the AI assistants where consumers actually engage.

UCP faces similar adoption questions. While backed by IAB Tech Lab’s established governance infrastructure, the protocol emerged just weeks after AdCP’s launch, adding to ecosystem fragmentation concerns rather than resolving them.

Expert opinion reveals deep divisions

Industry analysts offer sharply divergent assessments. 

Independent analyst Karsten Weide of W Media Research estimates 70% success probability over two to three years, emphasising AI’s momentum and open-source advantages. He defines success as more than half of major players adopting AdCP, with rapid developer adoption through easy integrations as the key driver.

Ruben Schreurs, CEO of Ebiquity and an AdCP launch member, provides a sobering counterpoint: “My estimation is this has a 20% to 30% chance to succeed in terms of getting to critical mass and scale.” He notes advertising’s history is filled with standards promising transformation but ultimately fragmenting or failing.

Veteran ad tech executive Ari Paparo offers the most nuanced verdict: brilliant for creative automation, deeply problematic for media buying.

On the creative protocol, Paparo sees genuine workflow improvements — standardising how AI agents handle assets across platforms addresses real complexity advertisers encounter. However, he expresses significant reservations about media buying, noting he “tried to build this at Google back in 2010 and ran into hostility and indifference from both sides.”

Paparo identifies four fundamental obstacles AI agents don’t necessarily solve: Buyers resist being “price takers” forced to accept published rates, sellers refuse to reveal rate card pricing, buyers don’t want the overhead of dealing with small publishers, and most publishers lack the scale to deliver performance. 

His conclusion: “The value will likely accrue to very large publishers and cross-publisher ad networks in much the same way it has with programmatic more generally.”

If you’d like to subscribe to my bi-weekly newsletter, INMA members can do so here.

Banner photo: Adobe Stock Albaloshi.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .