Peter Mason, illuma CEO, looks at how AI is reshaping publisher deals. He expands on how advertisers can find new audiences at scale, contextual AI, and the future of deal-making.

As 2026 planning cycles accelerate, one pressure point remains front of mind for advertisers and publishers: how to find new audiences at scale without losing relevance and performance.

For years, programmatic trading has forced a compromise. Buyers could run tightly defined private marketplace (PMP) deals with strong quality signals but limited scale, or open up reach and risk diluting relevance and performance. Meanwhile, the rise of premium video and CTV has made the equation even harder: more environments, fewer dependable signals, and increasing pressure to prove outcomes. 

That trade-off is now being challenged by a new model of deal-making – one where AI helps both sides to learn from what’s working in real time, react and scale to find relevant new audiences based on these signals. 

From fixed PMPs to living, learning deals

Traditional PMPs rely on static logic – predefined lists of audiences, sites and sections, or contextual categories, all locked in at campaign launch. They’ve been useful for buyers to a point, but they’re ultimately focussed on control, not scale, and this creates a plethora of challenges.  

Contextual AI is offering a next-generation alternative. Instead of depending solely on pre-set criteria, human-trained AI algorithms analyse content consumption during live campaigns, understand shifting audience interests and expand intelligently across publisher inventory. A live feedback loop simultaneously scales and optimises using buy-side performance signals so that audience relevance is maintained at peak throughout the campaign.

This process of real-time expansion delivers bespoke, qualified reach across publisher inventory and is particularly interesting for brands wanting to expand their audience base without relying on cookies or persistent IDs. Where historical lookalike modelling generally involves gathering past insights, they can now process content consumption signals as they happen. 

This means high-value seed audience behaviours are constantly modelled in flight, specifically when ad awareness is high, delivering vastly improved audience relevance and campaign outcomes. Crucially, the inputs are contextual, not ID-based, so personal data is not viewed, processed or stored at any point. New users are reached because they have closely aligned contextual interests to the seed audiences, in the same moment, irrespective of their profile.

Both scale and performance

A recent test between News Corp (publisher) and SAS (advertiser) shows what AI-powered audience expansion looks like within a PMP framework.

By allowing illuma’s contextual AI to guide expansion beyond narrow sections or predefined silos, the Smart Deal™ unlocked a much wider breadth of News Corp’s inventory and surfaced high-value content and audiences that might otherwise have been underutilised. Crucially, this was not scale at the expense of outcomes.

Across the test, SAS saw on average 3x performance uplift on News Corp properties versus other direct publisher activity not using illuma.

This combination of scale and efficiency has historically been difficult to achieve within PMP structures.

SAS global head of programmatic and direct buys, Mibbie Plouvier, said: “This new-found ability to grow our audience intelligently within a trusted publisher relationship was incredibly exciting for the programmatic team here at SAS. Contextual AI allowed us to target new audiences in relevant, high-quality environments while improving performance. It’s a very different way of thinking about publisher deals, and we’re excited to take it across platforms next; modelling our known audiences to find similar moments in video and CTV.”

For News Corp, the shift was equally meaningful. Rather than concentrating spend into a limited set of predefined sections, the test opened up opportunities across the full breadth of its highly valuable content. 

News Corp Director, ad tech programmatic strategy, Vikesh Chevli, said: “Smart Deals™ allowed our inventory to be evaluated in a more dynamic and intelligent way. Instead of being confined to static contextual definitions and pre-set audiences, our environments could prove their value through live performance signals. That’s a major innovation for our advertisers and a powerful shift in the way publishers will think about scaling deals in the future.” 

The future of deal-making

While this test ran within a defined publisher environment, the implications are broader. AI-driven audience modelling and contextual expansion logic should prove interesting for audience growth in more complex channels, including video and CTV, where traditional identifiers are even more constrained.

As media becomes more fragmented, the ability to understand which contexts are driving outcomes right now – and where similar signals exist – offers a pathway to expand reach without sacrificing relevance, with publishers sharing in the value of this optimisation.

In that future, deals don’t just transact impressions – they actively work to unlock better outcomes at scale.