Ranking
Ranking: How Search Results Get Their Order
Ranking is the stage everyone argues about and almost nobody can observe directly. This guide separates what is documented from what is inferred: how candidates are retrieved, which broad signal families do the scoring, and why the same signal matters more for one query than another.
Ranking is two operations that are often collapsed into one. Retrieval finds the set of documents in the index that could plausibly answer the query. Scoring puts that set in order. Retrieval is a fast, mechanical lookup; scoring is where judgement lives.
Retrieval narrows the field
A query for two terms becomes an intersection of two posting lists in the inverted index. From billions of documents, this reduces the candidate set to something manageable — typically thousands — in a few milliseconds. Query processing has already happened by this point: the query has been parsed, possibly corrected, expanded with synonyms, and interpreted for intent.
Scoring weighs signal families
Public documentation from Google describes broad families of signals rather than a formula. The honest summary is that the following all contribute:
- Meaning of the query. What the user is actually asking for, including intent, language and whether freshness matters for this kind of question.
- Relevance of the content. Whether the page contains the terms and concepts the query is about, in a way that answers it rather than merely mentioning it.
- Quality of the content. Signals of expertise and trustworthiness, including how other sites reference the page.
- Links. Which pages link to it, from where, and with what anchor text.
- Freshness. Weighted heavily for news and events, close to irrelevant for a definition that has not changed in a decade.
- Usability. Whether the page loads, works on the device being used, and is served securely.
- Context. Location, language settings, and in some cases prior activity.
The weights are private and query-dependent
Two things are true at once and both matter. The exact weighting is not published, and it is not a single fixed set of numbers — the same signal carries different weight for different queries. Freshness dominates for “election results” and is nearly meaningless for “what is an inverted index”.
This is why claims that a particular factor is worth a specific percentage should be treated as marketing. Nobody outside the engine has the weights, and there is no one set of weights to have.
What you can actually observe
You can observe your own rankings over time, the results a query returns, and the correlation between changes you made and changes in position. You cannot observe causation from a single site’s data. Where this site makes a claim about ranking, it either cites the engine’s own documentation or presents it as an observation with the method described.
Articles in this guide
The articles for this guide are being written. The guide above covers the whole topic in outline, and the article index lists everything published so far.