Attention prediction moves before the buy

Attention measurement has, so far, mostly been a rear-view mirror: you run the campaign, then find out how much it was watched. Amplified Intelligence wants to move the signal forward. Its AttentionAI, launched in June 2026, predicts how a given creative will be attended to before it is bought — an API-delivered, pre-flight attention estimate built on more than a decade of research and 50 billion-plus real human attention data points. It is a cross-channel product, not a DOOH tool. But the shift it represents — attention as a planning input rather than a post-campaign report — matters most for the media that can least afford to learn by spending, and beauty’s thin, cold-start networks are exactly that.

What happened

Amplified built its name on measuring actual human attention in the wild; AttentionAI turns that corpus into a prediction, so a planner can ask “how will this creative be watched?” before committing budget. The claimed pull is obvious for large advertisers running many variants — Connect’s ~500 tests a month is the shape of it. The number to keep at arm’s length is the 50 billion data points: it is a vendor-stated training figure, cross-channel, and says nothing beauty-specific. What’s genuinely new is the timing of the signal, and timing is where the value sits for anyone who can’t run a hundred tests to find out.

What it means for beauty

The structural problem of a beauty network is that it starts with no first-party attention data of its own — too few screens, too little history to know what creative will land. A cross-channel attention benchmark is a partial substitute; a predictive one is a better one, because it lets a small network reason about a creative in advance instead of burning scarce impressions to discover it doesn’t work. That is the same logic under the industry’s move to price on attention as the currency: if attention is what you’re selling, being able to estimate it before you ship is leverage.

The honest caveat is that this is AI applied to media, and the signal-versus-noise discipline still applies. A predictive model is only as good as its fit to your context — a salon mirror at close range in a long dwell is not the viewing condition most attention corpora are built on. Treat a pre-buy score as a planning prior, then validate it against your own outcomes, never as a delivered result.

The caveat that keeps us honest

AttentionAI is a single vendor’s cross-channel product; the training-scale and pilot figures are company-stated and directional, and none of them is a beauty or place-based number. What’s transferable is the direction of travel: attention is shifting from a post-campaign report to a pre-buy input, and that shift disproportionately helps media that can’t afford to learn by spending — which is the whole beauty category today. The Attention Unit and beauty economics stay modelled in the Research.