Speakers
Yu Jiang, Data & AI Practice Lead, Sia Partners
When a customer asks ChatGPT, Gemini or Claude which product to buy, the answer is your new storefront, and it is completely unaudited. LLM conversations leave no trace in your analytics stack, and customers are roughly 3x more likely to consult a third-party AI tool than your own chatbot. In this talk we show how we built a reproducible GEO measurement lab for one of the world's largest consumer brands: a pipeline that turns real customer language into synthetic personas and prompt banks, probes the major AI engines under controlled conditions, and scores every answer for accuracy, hallucination, share of voice and citation sources.
You'll see what the data revealed, including brands vanishing from unbranded queries, models coaching churn behaviour, and six different prices for the same question, plus how to move from measuring to experimenting: change one lever, re-run the lab, and prove what moved.
