An interesting new research paper just dropped: https://arxiv.org/pdf/2601.16858
It highlights fundamental differences between Google Search and generative AI systems.
Key takeaways:
• Once a document is included in an LLM’s context window (often influenced by SEO), its exact ranking matters much less for popular, high-coverage entities.
• For niche or low-coverage entities, ranking still has a huge impact on whether content is surfaced.
• Content freshness is critical in AI search ecosystems.
• Earned, trusted media sources strongly influence LLM responses.
This suggests GEO is not just “SEO for AI” it behaves very differently depending on entity maturity and authority.
Disclosure: I recently joined meikai.ai a week ago , where we analyze large-scale prompt data to understand how brands appear in LLM answers, and this aligns closely with what we’re seeing.
