James Dooley interviews Benjamin Tanenbaum from AISO about how personalization works inside large language models such as ChatGPT and Gemini. Benjamin Tanenbaum explains why AI search results change between users because models combine randomness, query fan out and contextual signals like location and conversation history. He shows that many differences in answers are not true personalization but distribution of possible responses, which means brands should measure visibility as share of voice rather than fixed rankings. The discussion matters because marketers must optimise content for LLM search even when answers vary, since stronger optimisation increases the probability of being cited in AI generated responses.