Every answer engine has its quirks, but the underlying optimisation model is converging. After six months of testing across three platforms, I've stopped writing separate strategies for each and started writing one.
What the engines actually reward
ChatGPT's web search, Perplexity, and Google's AI Mode all share three preferences: entity-dense content, explicit factual claims with sources, and consistent author attribution across the web. Sites that win citations rarely have the best individual pages — they have the most coherent overall presence.
- Cover the entity graph, not the keyword list.
- Make factual claims explicit and link to primary sources.
- Maintain consistent author bylines and bios across owned and earned media.
The unified framework
I now structure every piece of client content around four questions: What entity does this page own? What claims does it make? What evidence backs each claim? What other pages reinforce it? If a page can't answer all four cleanly, it gets restructured before publication.
The result, across the seven clients running this framework since November, is a 3-4× increase in LLM citations without a corresponding increase in word count. The pages aren't longer — they're more clearly authored.