• by Liam Lytton , Yesterday

When the chief strategy officer of the largest independent ad-tech platform leaves to build the money model for an AI lab, it reads as more than a career move. In May, Samantha Jacobson left The Trade Desk to lead monetization partnerships at OpenAI. The person who helped run the open web's biggest independent buying platform went to help an answer engine learn to sell, a fair signal of where attention -- and marketers -- are heading.

For two decades, the playbook held: pay for reach, rent the audience, collect the clicks. It still works when a business needs customers this month. But the ground is shifting. More people now ask a model instead of scrolling a page of links, and the model answers by naming businesses it has learned to trust. That rewards a different asset, which is not the audience you rent, but the recommendation you own.

Many teams are meeting softer campaign results by handing more control to automated bidding tools. It's an understandable reflex, and often the wrong one, because it treats a structural change as an optimization problem.

A large share of every ad dollar funds an opaque supply chain rather than durable brand equity, a dynamic underscored by this spring's transparency dispute between a major holding company and a leading platform, settled in June with little explanation (per at least one trade publication).

Here's what the move toward AI answers actually hands a brand: the model filters before anyone reaches the site. It reads the web, weighs trust signals, and points people toward what it recommends. By the time someone clicks, they're pre-vetted. That traffic behaves differently, converting better because intent is already formed.

That moves the bottleneck. When traffic is scarcer but higher-intent, web infrastructure becomes the main point of failure. Send an expensive, pre-qualified buyer to a slow page on a heavy template, and you burn the advantage you just earned. In a high-volume world, a three-second delay was tolerable. In an answer-first economy, it's a bounce. The fix is engineering: lighter pages, cleaner code, faster server response, and a path from landing to purchase quick enough that a ready buyer never feels a lag.

The strategy that follows runs on two tracks, because a brand now writes for two readers: people, and the crawlers that feed the models. The first is semantic authority, clean, well-structured data and clear schema that let an answer engine parse and cite a brand with confidence. The second is conversion speed; once the engine sends a qualified buyer to a brand's own domain, that domain must be fast and precise enough to close.

This rewrites the scorecard. Judging teams on sessions and impressions alone misreads how people now find things. The brands that do well will step back from spending that no longer compounds, and invest in what they actually keep. The same foundation that earns a ranking earns the citation.