AI Is Optimising Ad Campaigns Using Data No One Has Checked in Years
The sell side of advertising spent a decade building trust infrastructure. The buy side never built an equivalent, and AI agents are now exposing the gap.
Everyone building AI agents into advertising is asking the same question: what can this agent do. Almost nobody is asking the question underneath it: what is it learning from.
The argument getting attention
A recent industry op-ed put a specific, uncomfortable number on the problem: much of the buy-side data now feeding AI-driven marketing decisions, consent records, suppression lists, lead-scoring inputs, preference data, is years old and was never re-verified after it was first collected. The argument is straightforward. The sell side of advertising, publishers and exchanges, spent the better part of a decade building verification infrastructure: standards that let a buyer confirm who is actually selling an ad slot and whether the chain of custody is legitimate. The buy side never built an equivalent discipline for its own data.
That is not entirely true in the sense that no standard exists. A buy-side verification format was actually published years ago by the industry's own standards body. What did not happen is adoption. The standard exists on paper. Almost nobody enforces it in practice, which in some ways is a more uncomfortable finding than "nothing exists at all": the industry already agreed on how to fix this and simply didn't.
Why this matters more with agents in the loop
A human marketer working from stale data will eventually notice something is off, a segment performing strangely, a suppression list letting through people who unsubscribed years ago. An AI agent optimising a campaign at machine speed does not pause to wonder if its inputs are trustworthy. It just optimises harder against whatever it was given, which means bad data does not just produce a bad campaign anymore. It produces a bad campaign executed with total confidence and no human in the loop to catch it.
The Ad Tribe angle
This has not yet become the industry-wide reckoning the framing suggests, it is currently one sharp argument from one credible voice rather than a chorus. But it is the right argument at the right moment. The industry spent years teaching buyers to demand proof before they trust a seller. It has not yet asked buyers to hold their own data to the same standard, and agentic advertising is about to make that gap very expensive.