What is reranking?

By the Variask team · Updated

Definition

Reranking is a step in AI search where retrieved pages are scored again to decide which are most relevant to the question, before the model reads them. Pages that answer the exact question clearly tend to rank higher at this step.

How it works

  • A search step retrieves many candidate pages or passages.
  • A reranking model scores each against the question.
  • The top results are passed to the language model.
  • Only those can shape the answer or be cited.

What it means for brands

  • Clear passages that answer one question well survive reranking.
  • Pages that bury the answer are often dropped.

Questions

Reranking:
common questions

No. AI engines do not publish them. Track which pages are cited instead.

It is a similar idea applied to passages for an answer, often after a normal search.

Related terms

Keep exploring

How AI search works

Retrieval-augmented generation (RAG)

Retrieval-augmented generation (RAG) is a way of building AI answers where the system first retrieves relevant documents or web pages, then has the language model write the answer using them.

Read

How AI search works

Content chunking

Content chunking is the splitting of a page into smaller passages, so an AI system can store, find and quote the most relevant part instead of the whole page.

Read

How AI search works

Semantic search

Semantic search is search that understands the meaning and intent behind a query, not only its exact words, so it can return results that answer the question even when they use different wording.

Read
Full glossary