The Traitor in Your Media Plan

By Keith Arrowsmith, Marketing Director, Onetag

The return of The Celebrity Traitors puts trust back in the spotlight. Contestants must work out who is telling the truth while the people deceiving them try to stay hidden. In programmatic advertising, the stakes are different, but there is a similar challenge: knowing which signals deserve your confidence and which are leading you in the wrong direction.

As AI takes on more responsibility for media decisions, this question becomes increasingly urgent. The latest models process information at a scale no human team could match and adjust campaigns in real time. Speed, however, does not guarantee a good outcome. A model fed misleading signals will back the wrong inventory with complete confidence, much like a contestant who votes decisively for the wrong person.

The traitor hiding in plain sight

The best Traitors never look like Traitors.Poor-quality inventory does not always announce itself either. An impression can look cheap and a placement can clear the basic viewability threshold, but that doesn’t tell an advertiser whether the ad will deliver meaningful value. A Traitor who survives to the final takes the whole prize pot, and poor-quality supply that goes unnoticed takes a budget the same way.

Contestants also make the opposite mistake. Faithfuls are regularly banished on instinct, because someone seemed quiet or a story did not quite add up. Buyers do the same when they optimise towards a convenient proxy for performance, such as the lowest CPM, because the proxy is easy to see and the quality of the advertising experience is not.Programmatic spreads its information across many systems, so buyers struggle to see what sits behind an opportunity and how the inventory reached them.

Examining the impression in greater depth – the Traitor with a keener eye – closes that gap. Placement characteristics such as the context of surrounding content and attention explain why one opportunity is worth more than another, and live performance data shows how inventory behaves once a campaign is running. With both in view, advertisers can tell genuine value from an attractive-looking distraction.

A fast vote is not a better vote

AI suits this problem because it can interpret combinations of signals at a scale manual analysis cannot sustain. The system can weigh how different characteristics relate to campaign objectives and refine decisions as conditions change, not relying on just one metric. Inventory that performs gains priority, and placements that underdeliver lose it.

A fast vote is still a vote on whatever evidence the room has.If data is incomplete or its source unclear, automation repeats those weaknesses at greater speed. The quality of the information entering the system matters as much as the intelligence interpreting it.

This matters more as the industry moves towards agentic AI, where systems carry out tasks and adapt their decisions within a defined remit. IAB Tech Lab’s work on Agentic Advertising Management Protocols reflects the industry’s effort to make these workflows interoperable and governed by established advertising standards.

Trust the agent, but set the rules

Part of what makes The Traitors playable is that every player knows the rules: murders happen at night, banishments happen at the Round Table and the game is kept inside those lines. An agent needs the same clarity.Giving AI more autonomy does not mean handing over the campaign. Marketers still have to define what success looks like and how much freedom an agent has to act without approval, and those guardrails should cover every decision the agent can make, from selecting inventory to adjusting spend.

Human oversight matters most when a decision falls outside agreed parameters or the evidence behind it is thin. A person does not need to approve every routine adjustment, but consequential decisions must stay open to review and escalation. Transparency makes that possible. Advertisers need to see how their objectives become action, what informs each decision and where the budget goes. An agent that is efficient but cannot explain why a campaign performs as it does has still failed the marketer.

Keep the Faithfuls in the game

Agentic advertising connects decisions that have traditionally been fragmented across planning, activation and optimisation. When media quality signals and outcome data inform the same process, AI has a stronger basis for deciding where spend can work harder and publishers gain as quality environments are recognised for the value they offer. Banishing the wrong Faithful hands the advantage to the Traitors, and cutting a strong publisher because it looks expensive does the same for poor-quality supply.

In The Traitors, guesswork alone does not identify who is working against the group. Contestants need evidence they can interpret, and marketers face a similar task as AI takes on more of the work. The test is whether AI agents have the right information and context, and clearly defined limits, to make decisions worth trusting. Build those foundations and marketers can give AI more room to act without losing sight of what their media plan is meant to achieve.