Definition
Semantic search is retrieval that matches the meaning of a query rather than its literal words, so two different phrasings of the same question return the same thing. No standards body defines the term, and search engines rarely use it about themselves. It describes a direction of travel rather than one feature.
Example. Someone searching "why is my site not showing up on google" and someone searching "page not indexed" want the same answer, and share almost no words.
You can see the different phrasings AI search breaks a question into with the free Query Fan-Out Tool.
What has Google actually shipped?
Google has shipped a series of named systems that moved Search from matching words to matching meaning, and the dates are worth keeping straight, because most write-ups get the first one wrong.
| System | When | What Google said it did |
|---|---|---|
| Hummingbird | August 2013 | A major improvement to the overall ranking systems. Note: it was disclosed at a press event, and there is no Google blog post announcing it. Google's own record dates it to August, not September |
| BERT | October 2019 | Better understanding of one in ten searches in the United States, in English, at launch |
| MUM | May 2021 | Built on the T5 framework, described as 1,000 times more powerful than BERT, trained across 75 languages |
How does semantic search work?
Semantic search works by turning words into numbers that place similar meanings near each other, then retrieving by closeness in that space rather than by exact matches. Recognised entities do the rest of the work, which is why a well described brand is easier to retrieve than a well repeated keyword. Our entity SEO guide covers that side.
When does keyword matching still beat it?
Keyword matching still wins when the exact string is the intent. A part number, a SKU, an error code, a person's name or a quoted line of text all need the literal characters, and a meaning-based match returns something adjacent and useless. Good search systems run both and blend the results, which is why your product pages still need the exact terms printed on them.
Sources
- Google Search Central: a guide to Google Search ranking systems
- Google: understanding searches better than ever before, BERT (25 October 2019)
- Google: introducing MUM (18 May 2021)
Last checked against these sources on 22 September 2026.