RankBrain is a machine-learning system inside Google Search that helps it understand what a query means and how words relate to concepts, so it can return relevant pages even when they do not contain the exact words searched. Google first described it publicly in October 2015, and it still appears in Google’s own list of the ranking systems it uses.
How RankBrain works
Before systems like RankBrain, search engines leaned heavily on matching the words in a query to the words on a page. That breaks down for searches Google has never seen before, for long conversational phrasing, and for queries where the useful answer uses different vocabulary from the searcher. Someone typing “the tube strike thing that changed my season ticket refund” needs pages about rail ticket refunds during industrial action, even though few of those pages contain that sentence.
RankBrain represents words and phrases mathematically, so that terms with related meanings sit close together. When a new or unusual query arrives, it can relate it to queries and concepts Google already understands and help other ranking systems pick out pages that answer the underlying need. Google describes it as one of several AI systems in Search. Later ones do different jobs: BERT, for example, reads how word order and small words such as “to” or “without” change the meaning of a query.
A Google researcher told Bloomberg in 2015 that RankBrain was the third most important signal in ranking. That remark was widely repeated, but Google has never published a weighting, and it is better to think of RankBrain as part of how Google interprets queries than as a separate signal you can target.
Why it matters
RankBrain is one of the reasons repeating an exact phrase stopped being an effective tactic. Google matches pages to the meaning behind a search, so a page that covers a topic clearly in natural language can rank for many phrasings it never uses word for word. For a UK business, that means a well-written page about “boiler servicing in Bristol” can also be found by people searching “annual gas boiler check” or “how often should I get my boiler looked at”.
It also means the obvious route is the right one. Pages written to match search intent, in the words your customers use, and covering the topic properly are the pages these systems are designed to reward. This is the shift often described as semantic search.
Common mistakes
- Trying to “optimise for RankBrain”. There is no RankBrain setting, tag or checklist. Advice claiming otherwise is selling something.
- Believing it is driven by click-through rate or dwell time. This theory circulates widely, but Google has not confirmed it. Write for people because that works, not because of a rumoured metric.
- Writing one page per phrasing. Separate pages for “SEO consultant London”, “London SEO consultant” and “SEO consultancy in London” compete with each other. Google already understands they mean the same thing.
- Thin pages built around a keyword. A page that repeats a phrase without answering the question behind it has little for these systems to match.
How to act on it
Treat RankBrain as a reason to write clearly rather than as something to manage. For each important page, decide the single need it serves, then check the current results to see how Google interprets that need. Cover the topic fully, using the language customers use in enquiries, and answer the related questions a careful reader would have. Remove or merge pages that target near-identical phrasings.
Check Search Console regularly to see which queries a page appears for. If it shows for searches you did not target but which share the same intent, the page is being understood well. If it appears for the wrong intent, the content needs sharpening. Reworking existing pages in this way is the core of my on-page SEO work.