PurchaseDecision — Causal Measurement for AI Search

Problem

AI search is reshaping how consumers decide what to buy, but every AI-visibility tool on the market measures presence — citation share, sentiment, share of voice — not effect. A brand can know it is being mentioned in AI Overviews or ChatGPT answers without knowing whether those mentions change a single purchase decision. “Are we showing up?” and “did showing up change anything?” are different questions, and only one of them moves a budget defensibly.

Approach

PurchaseDecision runs randomized, single-session forced-exposure experiments. Real in-category shoppers are randomly assigned to test or control, shown pixel-realistic stimuli — Google results, AI Overviews, answers from ChatGPT, Gemini, and Perplexity, retail media placements, video ads — and measured before and after on awareness, favorability, consideration, and purchase intent, ending in an observed, screenshot-validated purchase decision. A statistics engine computes test-vs-control lift on matched, apples-to-apples groups (Coarsened Exact Matching or Inverse Probability Weighting, with logistic-regression lift models), so every study is analyzed the same way — consistent, repeatable, and comparable across studies.

Outcome

Live at purchasedecision.ai with a published proof-of-method experiment: 604 in-market dog food buyers, randomized, measuring what a brand mention in an AI answer actually did to the purchase decision. Studies go from brief to fielding in a day, with multi-market fielding across the Americas, Europe, Asia, and Oceania in local languages. Like the other Hubbs Enterprises brands, the platform was designed and built end-to-end by the founder.