Episode Details

James Dooley speaks with Sergey Lucktinov about how generative AI engines choose which sources to cite in AI answers. Sergey breaks down the full multi stage citation process, starting with query fan out, metadata filtering, and early result elimination, before explaining how LLMs deeply parse a small set of trusted pages. The discussion covers why prompt injection is risky, why semantic SEO and macro content structure matter, and how clarity, stability, speed, and cost of information retrieval influence whether a page is selected. They also explore why layout stability, time to first byte, and precise definitions increase trust, and how aligning content structure with how LLMs process information improves long term citation visibility.