GEO Experiments Show Third-Party Sources Drive 85% of AI Engine Citations

Two Generative Engine Optimization experiments conducted by Zeeshan Yaseen and reported by Search Engine Land reveal that roughly 85–86% of AI citations come from earned third-party sources such as reviews, editorial roundups, and partner pages, not branded owned content. The second experiment ran for 30 days across six AI engines — including ChatGPT, Claude, Gemini, Perplexity, Google AI Mode, and Grok — tracking 15 commercial-intent SaaS buyer queries. Researchers found that simply counting brand mentions in AI responses is an insufficient metric, as it does not indicate which buyer questions triggered visibility, which sources supported it, or whether it drove commercial results. The findings do not suggest companies should abandon their own websites, but warn that GEO strategies built primarily around self-published content may overlook the broader source ecosystem AI systems actually retrieve. Experts conclude that meaningful AI visibility measurement must be tied to specific buyer questions on specific engines, backed by identifiable third-party source surfaces.