How is a machine-readable ad deal structured for agent-to-agent negotiation?

A machine-readable ad deal is a structured data document containing the pricing terms, inventory specifications, targeting parameters, volume commitments, approval conditions and audit metadata agreed between a buyer agent and a publisher agent. It must be parseable by both agents without human translation, signed by both parties at commitment, and immutable after signing. The Agentic Advertising Management Protocols (AAMP), advanced by the IAB Tech Lab, defines the minimum field set. Deals that omit the audit metadata fields are not disputable through automated means.

A machine-readable ad deal is a structured data document containing the complete terms of a deal agreed between a buyer agent and a publisher agent, in a format both agents can parse and act on without human translation. It is the digital equivalent of a signed insertion order, but designed for autonomous agent consumption rather than human review.

At a glance

AAMP deal schema: The required fields in a machine-readable advertising deal under the IAB Tech Lab's Agentic Advertising Management Protocols: deal ID, buyer ID, seller ID, inventory specification, price terms, volume commitment, delivery period, and brand safety constraints.

Deal record: The committed version of the deal schema, written after both agents accept terms, held by the neutral marketplace and used as the reference point for compliance monitoring.

Compliance check: A query from the buying agent against the deal record comparing committed terms with delivery data — price, publisher environment, audience, completion rate — to identify discrepancies.

Bilateral deal ownership: The model in which neither the buyer nor the publisher holds the authoritative deal record alone; both access the same document from the neutral marketplace.

What fields must a machine-readable ad deal contain?

The Agentic Advertising Management Protocols (AAMP), under development at the IAB Tech Lab with a stable draft published in mid-2025, defines the minimum field set for a compliant deal record. A conforming deal document contains:

Deal identifier: a unique reference that both sides use to identify the deal across all subsequent communications.

Pricing terms: the agreed CPM or CPV floor, ceiling and committed price, along with the currency and any guaranteed volume or spend commitments.

Inventory specification: the publisher domains, app identifiers, content categories, placement types and environment parameters (desktop, mobile, CTV) to which the deal applies.

Targeting parameters: audience definitions, geographic constraints, frequency caps and any contextual targeting restrictions the buyer has specified.

Brand safety constraints: IAB category exclusions, viewability thresholds and any platform-specific brand safety requirements the buyer mandates.

Approval conditions: any steps that require human approval before the deal is activated, including spend thresholds above which human sign-off is required before additional budget is committed.

Commitment timestamps: the date and time at which each party committed to the terms, and the expiry conditions for the deal.

Audit metadata: fields that log any modifications to deal terms during negotiation, the agent identifiers of both parties, and a tamper-evident commitment confirmation from each side.

What makes a deal record "bilaterally verifiable"?

A bilaterally verifiable deal record is one to which both the buyer and publisher have independent access and which neither party can modify unilaterally after both have committed. Both agents retrieve the same document from the same source and receive the same content. If the record returned to the buyer differs from the record returned to the publisher, the marketplace has a governance failure.

This is the property that most current programmatic deal infrastructure does not provide. A DSP's deal record lives in the DSP's database. The SSP's deal record lives in the SSP's database. When those records diverge, which they routinely do on delivery and pricing details, there is no authoritative reference. Each side produces its own logs as evidence of what occurred. The dispute then requires human reconciliation.

Why does the audit metadata matter?

The audit metadata fields are the component most often omitted from early implementations of machine-readable deal records. They are also the fields that determine whether disputes can be resolved programmatically.

When a buying agent raises a delivery discrepancy, the resolution process requires knowing what was agreed, when it was agreed, by which specific agent versions on each side, and whether any modifications were made to terms after initial commitment. Without that metadata, a dispute resolution agent cannot determine whether the discrepancy is attributable to the publisher's delivery, a post-commitment term modification, or an interpretation difference between agent versions.

Deal records without complete audit metadata are adequate for executing deals but inadequate for resolving disputes. At the volume agents transact, the inability to resolve disputes programmatically is a material operational cost.

How does an agent use a deal record during a campaign?

During delivery, a buying agent periodically queries the deal record to confirm that inventory being delivered against the deal conforms to the committed specifications. If a publisher delivers inventory outside the targeted domains or below the committed viewability threshold, the buying agent raises a compliance flag against the deal record.

The flag triggers a defined escalation: automated query to the publisher agent, which checks against its own delivery log and responds. If both agents agree on what was delivered, the flag is resolved. If they disagree, the dispute is escalated according to the terms defined in the deal record's approval conditions field, which may involve a human reviewer or a neutral arbitration service.

This is the operational model that distinguishes a deal record designed for agent governance from one designed for human reference. The former is a live working document. The latter is a confirmation artefact.

What should buyers require from marketplace providers?

Buyers deploying buying agents should require that any marketplace they use produces deal records that conform to AAMP's field set, provides bilateral access (both buyer and publisher can retrieve the same record), immutability (records cannot be modified after both parties have committed), and query interfaces that allow buying agents to check deal compliance programmatically during delivery.

A marketplace that cannot demonstrate all four properties is not ready for production agent-to-agent deal deployment at scale. Alkimi's DealSheet architecture is one live implementation that provides bilateral access, immutability and programmatic query interfaces aligned to the AAMP field set.

Frequently asked questions

What is A2A protocol and how is it used in advertising?

A2A is Google's 2025 specification for agent-to-agent communication. In advertising, buying and publisher agents use it to exchange deal enquiries, counter-proposals, and commitments during deal negotiation, using AAMP to define the deal term structure.

What protocols let advertising agents transact directly?

A2A handles the messaging layer between agents; AAMP defines the machine-readable deal term schema both sides must implement. Together they allow a buying agent and publisher agent to negotiate, commit, and govern a deal without human intermediaries at the transaction level.

How does agent-to-agent negotiation reduce the number of intermediaries in ad buying?

Direct agent-to-agent deal negotiation via a shared marketplace replaces the chain of SSP, exchange, and DSP systems through which deal terms currently travel. The bilateral deal record held by the marketplace eliminates the need for separate deal management systems on each side.