The growth of large language model (LLM) platforms as consumer purchase research channels has generated a corresponding growth in AI visibility measurement tools. Platforms including Profound, Peec, Otterly, Siftly, and others track brand citation frequency, share of voice, and mention rates across ChatGPT, Perplexity, Gemini, Grok, and Claude. This measurement infrastructure addresses a genuine commercial need: brands require systematic data on how AI systems represent them. However, the measurement frameworks currently available share a structural limitation. They measure the awareness layer of AI-mediated consumer behaviour — the first-turn response to a category query — rather than the decision layer, where the consumer asks an AI model to make a purchase recommendation. These are different stages of the buying journey, and they produce systematically different competitive outcomes. This paper documents the AIVO Paradox: the empirical finding that high AI visibility does not predict, and frequently coexists with, low AI purchase recommendation win rate. A brand can appear in 9 of 10 category queries while winning the final buying recommendation in fewer than 10% of purchase-intent sequences. The gap between these two measurements — visibility and recommendation — is the AIVO Paradox. It is not a measurement artefact or a sampling anomaly. It is a structural feature of how LLMs process buying conversations, and it has material commercial consequences. We present the theoretical basis for this divergence, the empirical evidence across 137+ brands and five categories, and the three sub-types of the paradox that capture its distinct manifestation mechanisms. We also document a second divergence — between directed probe outcomes and undirected journey outcomes — that reveals a further layer of measurement inadequacy in current AI brand intelligence practice.
Tim de Rosen (Tue,) studied this question.