Brand safety in an agent-negotiated deal: who is accountable?
Sep · 5 min read
When an agent commits to inventory on a buyer's behalf, accountability is defined by the approval record and deal terms. Here is what buyers need in the deal structure.
brand safety
Nothing matches that. Try a different word or tag.
What Concourse Bench v1 reveals about AI buyer error rates
Sep · 5 min read
55 completions from 96 attempts. 95 of 110 contracts with sub-70% viewability floors. The first rigorous public data on how AI media buyers actually perform.
How to set the right boundaries for an AI buyer without killing its effectiveness
Sep · 5 min read
Which decisions should the agent take autonomously, which require sign-off, and how is the boundary enforced? A practical guide to the harness-level design question.
News
Brand Safety in Agentic Advertising: A Guide for US Buyers
Sep · 4 min read
Brand safety in agentic advertising requires buyers to encode safety constraints directly into agent mandates, verify that those constraints travel with the agent across every negotiation, and audit outcomes against declared parameters.
How Agentic Advertising Platforms Handle Invalid Traffic and Ad Fraud
Sep · 2 min read
Invalid traffic and ad fraud are persistent problems in US programmatic advertising. Agentic advertising platforms need to address IVT not only at the impression level but at the deal governance level, where agent-negotiated terms can be manipulated if ...
How to Maintain Brand Safety When AI Agents Are Buying Media
Sep · 4 min read
Brand safety in agent-mediated buying is maintained through the mandate, not real-time human review. The mandate must include explicit inventory exclusions and adjacency requirements. Post-campaign verification uses the DealSheet as the reference document.
Why Verification Vendors Are Moving Into Deal Record Auditing
Sep · 4 min read
Why verification vendors are expanding into deal record auditing: the three structural advantages they hold, what deal record auditing involves, and what buyers should ask their current verification vendors.
Brand Safety in Agentic Media Buying: What Agents Check and What Requires Human Sign-Off
Sep · 8 min read
Brand safety in programmatic advertising has always been a two-layer problem. Understanding the distinction between context safety and creative safety is the difference between deploying agents correctly and discovering the gap after an incident.
Insight
From MFA to Agent-Negotiated Inventory: Why Transparency Just Got a New Problem
Sep · 6 min read
MFA detection happens post-buy today. In agentic advertising, the deal is agreed before the buy, moving quality evaluation to the point of negotiation or turning post-buy detection into an audit of a fait accompli.
Natural-Language Confirmation Is Not a Transaction: The Verification Problem at the Heart of Agentic Advertising
Aug · 5 min read
IAB Tech Lab's Shailley Singh: agentic advertising cannot rely on natural-language confirmation. Research across 90,202 simulated deals found 95.3% disagreement between separate agent records, falling to 0.19% with a shared record.
Brand Safety in Agentic Advertising: What Verification Looks Like at Machine Speed
Aug · 8 min read
How brand safety controls work when buying decisions happen faster than human review cycles.
The Verification Layer that Programmatic Needs
Jun · 7 min read
Why an independent evidence layer, separate from ad servers and measurement vendors, changes what trust means in programmatic.
Measurement Isn't Verification: Fifteen Years of Self-Graded Essays
May · 6 min read
Why the industry's reliance on self-reported measurement has never been a substitute for independent verification.
Research
CTV viewability floors when the buyer is an AI: what the contract should say
Sep · 5 min read
95 of 110 AI-negotiated contracts had sub-70% viewability floors. In CTV, where standards are more stringent, that gap is a serious problem. Here is what the contract needs to fix it.
Three ways an AI buyer fails a media brief, and what each failure tells you
Sep · 7 min read
41 of 96 attempts in Concourse Bench v1 did not produce a completed buy. The failure modes are distinct, and the distinction matters for what you fix.