RRF · Reciprocal rank fusion
Reciprocal Rank Fusion (RRF) is a method that merges several rankings into one: each document gets a score based on its rank in each list. It is common in hybrid search and cited as a way to aggregate the results of a query fan-out’s sub-queries, though Google has not documented its exact mechanism.
Standalone definition · 53 words
How does RRF work?
For each list, a document’s score is 1 divided by (k + rank), with k set to 60 in the original paper by Cormack, Clarke and Buettcher (SIGIR 2009). Scores from all lists are added. Example: a document ranked 1st in one list and 3rd in another gets 1/61 + 1/63, about 0.0323.
Why does RRF matter for GEO?
A document present in several rankings rises, even if it is never first. If this principle applies to a query fan-out, it favors pages that answer several sub-queries well, hence the value of covering several angles rather than targeting a single keyword.
Where do you find it?
In hybrid search, which combines lexical and vector results, including in engines such as Elasticsearch or Azure AI Search. A site has nothing to configure: it is a principle that explains why multi-angle coverage pays off.
Frequently asked questions
Does Google use RRF?
Google has not published the detail of its fusion mechanism. RRF is a public, common method, which does not prove Google uses it.
What does k mean in the formula?
It is a smoothing constant, set to 60 in the original paper, which limits the advantage of the very top ranks.
Does RRF use scores or ranks?
Only ranks, which makes it possible to merge lists whose scores are not comparable.
Related terms
Further reading
- RRF, Reciprocal ranking fusion tech class the answers on the LLMs (Chatgpt, Gemini)
- How to track your position in google ai mode and ai overviews
- ← Back to the GEO glossary (114 terms)