What Is AdCP? AI Ad Creation, Approval, Delivery, and Corporate Responsibility
What you’ll learn
- AdCP’s role and the path from an advertising request through creation, review, and delivery
- Why proposing an ad, checking it against rules, and delivering it are separate roles
- What an AI-generation record or right-use assertion cannot establish on its own
- The checks required before delivery and how an organization can divide responsibility
What AdCP Handles and What the Organization Must Manage
Finishing an AI-created ad is not enough to make it ready for delivery. Someone must confirm that it follows the brand’s rules, that the organization has permission to use source images and a person’s likeness or voice, and that the approved version is the version actually delivered.
AdCP (Advertising Context Protocol) is an open standard for passing the information needed across creation, review, and delivery in a shared format. When an advertiser’s AI tools and advertising providers use the same format, they can carry review outcomes and the approved version forward even when the people or tools change.
Passing information is not the same as deciding that an ad may run. AdCP provides shared handoff procedures. It does not automatically validate usage rights or complete final approval.
This article explains AI advertising governance—the rules and management structure used to create, approve, and deliver ads appropriately—by separating what AdCP can handle from what an organization must decide and manage itself.
Shared Procedures from Creation to Delivery
AdCP is a protocol: a shared set of rules that systems use to exchange information. The specification and JSON Schema—the rules a computer uses to check field names, formats, and required fields—define the formal contract. The public server is an illustrative implementation rather than the specification itself.[1]
The basic workflow for advertising material is as follows.[2]
- Advertising request: State the objective and production conditions.
- Delivery-condition check: Retrieve the formats and constraints required by the destination.
- Ad creation or registration: Create material for those conditions or register existing material.
- Finished-preview review: Check how the ad will appear when published.
- Approved-version transfer: Send the approved version to the delivery side.
The official documentation also treats creation, finished-preview review, and transfer of the delivery version as separate operations.[2]
This procedure is useful when a team works with several creative tools or media destinations, such as websites, apps, or print publications. It lets the team check each destination’s dimensions and required information before the end of production, when differences are more expensive to resolve.
Linking the finished preview and the status of the approved-version transfer to the same management number also makes it easier to trace who checked what and which version was delivered.
A technically valid file can still violate brand rules, permissions, law, or local expectations. Someone other than the creator must decide whether it may be delivered.
Separate Roles for Proposals, Rule Checks, and Delivery
AdCP’s Governance documentation separates the role that proposes a campaign plan, the role that checks the plan against rules, and the role that fulfills an approved media transaction.[3]
| Role | AdCP term | Main responsibility |
|---|---|---|
| Advertiser-side coordinator | Orchestrator (the system that coordinates multiple AI tools and processes) | Proposes campaign plans and executes media transactions |
| Reviewer independent of the proposal | Governance Agent (the system that checks budget, brand, and organizational rules) | Evaluates the proposed plan or transaction and returns a decision |
| Advertising inventory and delivery side | Seller (the party that provides advertising inventory and delivers the ad) | Fulfills media transactions within the approved scope |
When a proposal does not meet the conditions, the process can reject it or route the decision to a person while recording the result.[3]
This separation prevents the AI system that creates the ad from also controlling spending and final approval. For example, an organization can permit an AI-assisted draft while stopping the following ads before delivery:
- wording or imagery that conflicts with brand rules
- an ad intended for a region outside the approved scope
- an ad that exceeds the budget limit
- an ad with no recorded permission for source material or a person’s likeness or voice
Separating roles does not by itself stop delivery. If AdCP procedures are not connected correctly or are misconfigured, a seller may transact without calling the independent check.[4]
The deploying organization therefore needs a mechanism that actually blocks a rejected ad on the delivery side. It also needs internal rules, human approval, and post-delivery review.
AI-Generation and Right-Use Assertions Still Require Separate Checks
AI-Generation Records Are Not Creation Histories
In AdCP 3.1.2, the version reviewed for this article, an advertising payload can include a Boolean value—a two-state value such as yes or no—showing whether an image was AI-generated, as well as a right-use assertion.[4]
A record that says “AI-generated” does not identify who created the image, which AI they used, when they created it, or how it was edited.
The official documentation for this version states that AdCP does not provide a cryptographically signed creation history that accumulates assertions across generation, editing, and adaptation. Interoperation with C2PA, a technical standard for recording the source and editing history of digital content with signed information, is future work.[4]
A Right-Use Assertion Is Not a Rights Review
A right-use assertion means that the sender says it has the right to use the material. The assertion alone does not establish that a contract or a person’s consent is valid for the planned use.
The same boundary applies to a rule-check result. Passing configured rules does not establish compliance with every law or every advertising provider’s rules.[4]
Consumer-Facing AI Disclosures
AdCP 3.1.2 does not define a common mandatory disclosure requirement for AI-generated content in the delivered ad.[4]
After receiving the AI-generation record, right-use assertion, and rule-check result, advertising providers and deploying organizations must still evaluate the validity of contracts or consent, required disclosures, and applicable law for the relevant region and use.
The following table separates what the received information indicates from what it cannot establish on its own. This is an editorial classification for this article, not an official AdCP classification.
| Information received | What it indicates | What it cannot establish on its own |
|---|---|---|
| AI-generated image record | The sender recorded the image as AI-generated | Who created it, which AI was used, how it was edited, whether the history is complete, or whether the record is signed |
| Right-use assertion | The sender says it has the right to use the material | Whether a contract or consent is valid |
| Rule-check result | The result of checking defined rules | Compliance with every law and every advertising provider’s rules |
Pre-Delivery Checks and Organizational Responsibilities
Before delivery, the organization must decide who reviews the handed-off information and what evidence supports each decision.
The following AI Advertising Governance Handoff organizes the information, reviewer, and check for one advertising project. It is an operational proposal from this article, not an official part of the AdCP specification.
The tables include information carried by AdCP as well as information that an organization stores or references independently. Not every entry is a standard AdCP field.
Ad Identification, Creation, Brand, and Rights
| Check | Information handed off | Owner and review | Problem to prevent |
|---|---|---|---|
| Ad management number | Campaign, creative, version, and destination | The delivery-records owner confirms that the same management number and approved version are used from review through delivery | Delivery of a different ad or a version changed after approval |
| Creation | Request describing the advertising objective, format, finished material, and finished preview | The creative owner confirms that only approved AI services and permitted input information were used | Use of an unapproved service or prohibited input |
| Brand | Brand name and applicable rules | The brand owner reviews the claims, wording, imagery, and likely reception in the target region | Misleading, off-brand, or locally inappropriate advertising |
| Rights | Right-use assertion and location of the contract or consent record | The legal or rights owner matches contracts and permission to the source material, person’s likeness or voice, and intended use | Use of material or personal information without permission |
AI Use, Human Approval, Disclosure, and Delivery Records
| Check | Information handed off | Owner and review | Problem to prevent |
|---|---|---|---|
| AI-use information | How AI was used, who made the assertion, references to C2PA or embedded creation information, and review results | The AI-use policy owner decides which creation information its policy requires and how it will check the assertion independently | An unsupported account of how the ad was made |
| Human approval | Governance result—the result of checking the defined rules—and any escalation to a human decision owner | The business owner records the approver, covered version, approval end date, substitute approver, and stop conditions | Reuse of expired approval or approval for another version |
| Consumer-facing disclosure | Declared requirement, region, wording, and display instructions | Legal and advertising-provider-rule owners decide the wording, placement, region, and destination | A required disclosure is missing or shown incorrectly |
| Delivery record | Advertising provider, approved-version transfer status, and decision record | The delivery and records owner saves what was delivered and records any later stop or removal | An ad cannot be traced or removed |
At delivery time, review the ad management number, approved version, rights review, rule-check result, and consumer-disclosure decision as one set. If any part changes, require another approval and make sure an older version does not remain in the delivery queue.
Scope and Checks for an Initial Deployment
An initial deployment is easier to evaluate when it is limited to one advertising project, one ad format, and one advertising provider.
The organization should confirm that the same management number connects the advertising request, finished preview, brand review, human approval, approved-version transfer, and post-delivery record. It should also confirm that information can make a round trip between the creation and delivery sides.
Three checks are especially important:
- Enforcement of a stop decision: When the independent reviewer requests a stop, does the advertising provider actually prevent delivery?
- Approval-expiry handling: When approval expires, does the process require another review?
- Review after material changes: When the ad is replaced, is the changed version reviewed again?
Even a correctly structured AdCP message is not an effective control if the delivery side does not act on the decision. An initial deployment must test not only the information format but also whether approval and stop decisions change actual delivery.
Summary: AdCP Does Not Remove the Organization’s Decision Responsibility
AdCP provides shared procedures for carrying information needed for AI ad creation, finished-preview review, rule checks, and delivery across different people and systems. It makes it easier to trace which ad was reviewed, who reviewed it, and which version was delivered.
The deploying organization and advertising provider still own decisions about whether contracts and consent are valid, which consumer-facing disclosures applicable law and destination rules require, how internal rules are enforced, and who gives final approval. An AI-generation record, right-use assertion, or rule-check result does not prove a complete creation history, validated rights, or legal compliance.
The practical distinction is among information carried through AdCP, conditions that the organization’s systems enforce, and decisions that people make. Applying the handoff tables to one advertising project helps expose the additional checks and responsibilities the organization must define.
This article is a general information summary and is not legal advice. Confirm practical decisions with a qualified specialist.
References
The specification discussion covers AdCP 3.1.2, the version reviewed for this article.
- Ad Context Protocol, AdCP - Advertising Context Protocol
- Ad Context Protocol, Creative Protocol
- Ad Context Protocol, Governance Protocol
- Ad Context Protocol, Trust & Security
For the latest releases and updates, check the official website and official documentation.