ChatGPT doesn’t just answer questions - it ranks info using math. And now we know how: it uses something called RRF.
People don’t realize how AI picks what to show in answers. This article uncovers the math behind ChatGPT’s ranking system and shows how websites can win more AI visibility.
ChatGPT uses a formula called Reciprocal Rank Fusion (RRF) to decide what results to show in answers. RRF gives small scores to links based on how high they rank in different searches, then adds up those scores. So, if your page ranks in many related searches, even if not always at the top, it still scores better than a page that only ranks #1 for one search. This is great news for websites that cover full topics in depth instead of chasing just one keyword.
The article proves this by showing examples in the code from ChatGPT’s dev console. It explains how AI tools like ChatGPT and Perplexity run many searches at once and combine them using RRF. The more places your content shows up, the better. The article also shows how topic clusters - a main page plus many subpages - are perfect for this system. The more related queries your site can answer, the more RRF points you get, and the more likely AI will show your content.
In short, search is now about being consistent and useful across a full topic, not just winning a few big keywords. If your site is seen as an expert on a topic, AI search engines will reward that.