Advertising Finally Has a Common Language with AdCP 3.0

By: AgenticAdvertising.Org

Today, AgenticAdvertising.Org publishes Ad Context Protocol 3.0 — the most significant expansion of AdCP since we launched the standard six months ago, and the clearest signal yet that agentic advertising has moved from experiment to infrastructure.

What is AdCP — and why does it matter?

Before we get to what's new, let's talk about what AdCP actually is because "protocol" is one of those words the industry tosses around without always explaining.

A protocol is a shared set of rules that lets different systems talk to each other. HTTP is the protocol that makes websites work across every browser and device. OpenRTB is the protocol that made programmatic bidding possible across thousands of platforms. You don't think about either of them most days, you just benefit from the fact that they exist.

AdCP is the protocol that makes agentic advertising possible. It is the common language that allows AI agents — across the whole value chain of advertising — to discover inventory, plan campaigns, execute buys, generate creative, govern decisions, and report results across any platform, any format, any channel. Without a shared language, every integration is custom. Every new partner is a new project. There is no interoperability and agents can't collaborate if they can't communicate.

AdCP 2.0 established that language for media buying and signals. AdCP 3.0 extends it to the entire advertising lifecycle.

The shift we're describing

The advertising industry has spent twenty years optimizing auctions. Programmatic made it possible to bid on billions of impressions in milliseconds. It was a remarkable feat of engineering but it left a lot on the table.

It never gave brands access to the inventory they actually wanted. It never let publishers sell their best units at the value they deserved. It never built the infrastructure for forecasting, for guaranteed buys, for unified reporting across channels. It was designed to optimize the transaction, not the relationship.

AI agents change what's possible, but only if the infrastructure exists to support them. An agent that can plan, buy, and optimize across every channel in real time is transformative. An agent that has to speak twelve different languages to do it is just complexity disguised as progress.

AdCP 3.0 is the infrastructure that makes agents genuinely powerful. It gives every participant in the ecosystem — publishers, brands, agencies, platforms — a shared foundation to operate from. One protocol. Every platform. Every step of the campaign.

What's new in AdCP 3.0

AdCP 3.0 is not an incremental update. It is a structural expansion of what the protocol covers, moving from media execution into the full lifecycle of advertising — discovery, planning, execution, governance, identity, and creative. Here are the capabilities that define this release.

Governance That Is Architectural, Not Optional

The question the industry keeps asking about agentic AI is: who is responsible when an agent makes a decision? AdCP 3.0 has a structural answer.

We call the approach Embedded Human Judgment: humans stay in the loop on decisions with real consequences, not on every operational step. The protocol enforces this structurally. Before any campaign goes live, check_governance validates budget authority, brand safety, and regulatory compliance automatically. If an action exceeds a human's pre-approved authority threshold, the system escalates — the agent cannot proceed until a human approves. After every campaign, a complete audit trail documents every decision: who proposed it, who approved it, what conditions were attached, what actually ran.

For fair-housing, fair-lending, and fair-employment campaigns — categories where the EU AI Act's Annex III obligations apply — prohibited targeting combinations are rejected at the schema layer rather than the policy layer. The request is structurally invalid before the policy engine runs. Thirteen regulatory policies ship pre-seeded in the protocol's policy registry, covering US (FHA, ECOA, EEOC, COPPA, FDA, FTC, TTB, state gambling), EU (DSA, prescription DTC, GDPR Article 9 special categories), and platform-level restrictions.

This is what Embedded Human Judgment means in practice. Not a human watching every move — that would eliminate the efficiency that makes agents valuable. A system that catches what humans miss, enforces the limits humans set, and traces every decision back to its source. The people responsible for strategy and spend remain responsible. The agents do the operational work. That's the right division of labor.

Brand Identity That Travels With the Campaign: brand.json

For as long as advertising has existed, brand identity has lived in PDFs. Brand books, style guides, asset libraries — documents that require a human to interpret and apply them, manually, every time a creative is produced or a campaign is planned.

AdCP 3.0 introduces brand.json: a machine-readable specification that encodes a brand's identity — visual guidelines, tone, colors, logos, licensing rights — into the protocol itself. Think of it as a robots.txt for brand identity. Any authorized agent — a creative tool, a buying platform, an AI assistant — can read it directly and apply it immediately. No PDF. No manual handoff. No interpretation errors.

Just as publishers use adagents.json to declare their properties and authorized agents, brands now use brand.json to declare who they are and what they authorize. Together, these two files establish provenance and trust on both sides of the ecosystem. Publishers and brands can both operate with confidence that every interaction is authorized and on-brand.

For brands managing creative at scale across multiple channels and AI generation workflows, this is foundational. Your identity travels with your campaign. Agents can generate on-brand creative from a single brief, verified in real time, at any scale.

Advertising Inside AI: The Sponsored Intelligence Protocol

Consumers are spending more time inside AI assistants. They're asking for product recommendations, comparing options, making decisions — all in conversation. Advertising has had no standard way to show up in these environments in a way that respects the user, the platform, or the brand.

AdCP 3.0 introduces the Sponsored Intelligence Protocol: the first open standard for how brands participate in conversational AI. It defines the full lifecycle of a brand session inside an AI assistant — consent, engagement, creative formats, commerce handoff — so that brands can show up in these environments in a way that is integrated rather than interruptive.

This is not banner ads for chatbots. It is a consent-first protocol where a user's expressed interest opens a structured session with a brand, enabling rich, contextual engagement that fits the conversational flow. The protocol handles everything: session management, capability negotiation, UI components, transaction handoff. It works across any AI platform because it's a standard, not a one-off integration.

The Creative Protocol: One Brief, Every Format

Producing creative for a multi-channel campaign today means six production workflows, six sets of specs, six rounds of asset management. It's one of the most persistent operational bottlenecks in the industry.

AdCP 3.0's Creative Protocol collapses this into a single workflow. A creative agent receives one brief, reads the publisher requirements from every connected seller, generates on-brand assets — pulling directly from brand.json — in every required format, verifies compliance in real time, and distributes to every seller simultaneously. What used to take weeks of back-and-forth takes one automated workflow.

The protocol also enables iterative creative refinement in natural language. "Make the opening more dynamic. Swap the product shot." The agent understands, adapts, and produces. Humans stay in the creative loop without being in the operational one.

20 Channels. One Protocol.

AdCP 3.0 expands from 9 to 20 media channels — including linear TV, radio, streaming audio, gaming, influencer, cinema, email, print, and sponsored intelligence. It also introduces planning infrastructure that programmatic never had: delivery forecasts, guaranteed buys, proposals with structured budget allocations, and a formal accounts protocol for managing commercial relationships across platforms.

This matters because it means AdCP can power buys that programmatic never could. Linear TV purchased through the same protocol as digital. Premium publisher inventory that never surfaces in an auction. Forecasting. Guaranteed delivery. AdCP-enabled sales agents are in production for platforms including Snap, Pinterest, Reddit, Netflix, and Vox Media, with more coming online weekly.

What AdCP 3.0 does not do

The credibility of a standard is measured by its gap list as much as by its feature list. AdCP 3.0 is explicit about what it does not address — documented at /docs/reference/known-limitations. A few to note:

AdCP operates at the campaign and deal layer. It does not replace OpenRTB for impression-level auction — AdCP agents hand off to OpenRTB infrastructure at the execution layer. AdCP Verified, the protocol's conformance tier, ships as a self-attested standard with a public test suite; independent third-party audit is on the 3.1 roadmap. In-protocol consent-scope propagation is not included in 3.0; user consent state travels through existing frameworks (TCF, GPP, and jurisdictional equivalents), and AdCP defers to them.

Naming these gaps is part of how we build a standard that deserves trust.

Who the AAO is — and how it's governed

AgenticAdvertising.Org (AAO) exists because advertising needed a neutral, open body to build and maintain the infrastructure for agentic AI. We are not a platform. We are not a vendor. We are an industry consortium — publishers, brands, agencies, technology companies — working together to build the standards that make the whole ecosystem work better.

Specific structure matters more than adjectives here, so here is the specific structure:

AAO seats its first elected board on May 6, 2026 at the first Annual General Meeting. The board has four voting classes — brands, agencies, publishers, and technology providers — with ten seats per class and equal voting weight across classes. Board members serve staggered three-year terms with a two-consecutive-term limit. An executive committee is drawn proportionally across classes. Until May 6, AAO operates under a four-director interim board with published recusal rules.

AdCP the specification is published under Apache 2.0 with its express patent grant. The sell-side reference implementation is maintained by Prebid — a software organization, not a standards organization — under its own community governance. The specification and the reference code are deliberately held by separate organizations.

Open standards are how advertising infrastructure has always been built. OpenRTB didn't belong to a DSP. VAST didn't belong to an ad server. AdCP doesn't belong to any one company. It belongs to the industry, governed by its members, and open for anyone to build on.

What to do next

AdCP 3.0 is public. The documentation, schemas, and implementation guides are open. The limitations page tells you what we haven't yet done. The governance structure tells you who gets to decide what we do next.

If you're implementing: start at docs.adcontextprotocol.org. The full protocol is public, the schemas are open, and domain-specific implementation guides are available for buyers, sellers, creative agents, governance agents, and data providers.

If you want to understand the concepts: the AAO Academy's Level 1 course takes about 90 minutes and earns a digital credential. It's the fastest way to build working fluency in the vocabulary. agenticadvertising.org/certification

If you want to shape where this goes: membership in AAO includes voting rights in the elections that seat the May 6 board. Membership levels and rights are at agenticadvertising.org/membership.

AdCP 3.0 is the infrastructure that makes the agentic era of advertising possible. It was shaped by people across the industry. It is released for the industry to use. The work of building on it begins now.