Half Your Media Budget Is Now Invisible to Your Audience

For decades, media planning operated on a simple model: audiences consume content on publisher platforms, and brands buy access to those audiences through advertising. Every metric — reach, frequency, CPM, viewability — was built around this model.

AI intermediation breaks the chain. The audience asks a question. The AI provides the answer. The publisher's content was used to train or inform the AI's response, but the audience never visits the publisher. Your ad was on the publisher's site, but the audience was never on the publisher's site.

The industry response has been to pretend this isn't happening. CPMs are still calculated on publisher traffic. Media plans still optimise for placement on publisher properties. Measurement still assumes a direct relationship between ad exposure and audience attention. It's like planning a newspaper advertising campaign in 2010 — technically you can still do it, but you're optimising for a distribution channel that's in structural decline.

The money hasn't followed the attention because admitting the shift means admitting that a significant portion of current media spend is waste. And nobody in the media buying chain — agency, ad tech intermediary, or publisher — is incentivised to admit that.

Attribution was already difficult. Multi-touch models, last-click attribution, incrementality testing — the industry spent years trying to solve the measurement problem without ever quite succeeding. AI intermediation hasn't made attribution harder. It's made it honestly impossible — which is paradoxically more useful than the false precision we had before.

When an AI agent recommends your product, what caused that recommendation? Was it your content strategy? Your brand authority? A specific piece of structured data? You can't A/B test an AI's training data. You can't track a cookie through a language model.

The useful part is that this finally forces the industry to stop pretending that last-click attribution was ever real. It wasn't. It was a convenient fiction that gave programmatic advertising a measurement advantage over brand building — not because programmatic was more measurable, but because the measurement was more precise-looking. The AI intermediation era exposes what was always true: most of what we thought we were measuring was correlation, not causation.

The opportunity is to rebuild measurement around what actually matters: brand authority, consideration set inclusion, and conversion from direct relationships. These are harder to measure precisely — but at least the measurement is honest.

Publishers are facing an existential threat from AI intermediation. If AI models summarise their content without sending traffic, the ad-supported publishing model collapses. We're already seeing this: publishers blocking AI crawlers, negotiating licensing deals, experimenting with paywalls.

For brands, this creates a problem that most media plans ignore. The content supply chain you depend on for media placement is degrading. Publisher audiences are shrinking. Quality journalism is being defunded. The environments where your ads appeared are getting worse — more programmatic filler, more AI-generated content, fewer real readers.

The brands that navigate this well are the ones investing in direct audience relationships rather than renting access to someone else's declining audience. Owned media, first-party data, community — the boring fundamentals that don't show up in a media plan but increasingly represent the only reliable channel to market.

The brands that don't navigate this are the ones still arguing about viewability thresholds on display ads that nobody is looking at.

Media planning now has two audiences: humans and AI agents. The human audience still needs to be reached through traditional and digital channels. But the AI audience — the models that increasingly mediate between your brand and your customer — needs a completely different approach.

Reaching AI agents means structured content, entity authority, and the kind of semantic infrastructure that makes your brand a credible source in AI-generated responses. It means investing in the same things that make GEO work: deterministic measurement, authority graphs, and content structured for machine comprehension rather than human browsing.

Most media plans don't account for this at all. They allocate 100% of budget to reaching human audiences and 0% to ensuring the brand is represented in AI-mediated discovery. That split made sense two years ago. It doesn't any more.

The uncomfortable question for every CMO is: what percentage of your media budget should be allocated to influencing how AI represents your brand? The answer isn't zero. And the organisations that figure out the right number first will have a structural advantage that compounds over time.