Hummingbird is the name Google gave to a rebuild of its core search algorithm, announced in September 2013, that moved search away from matching individual words and towards understanding the meaning of a whole query. It was not a penalty or a filter but a replacement for the engine itself.
How Hummingbird works
Before Hummingbird, Google leaned heavily on matching the words in a search to the words on a page. Hummingbird was designed to read a query more as a person would: to work out what the searcher meant from the whole phrase, including small words such as “near”, “for” or “without”, and to connect it to concepts rather than strings of letters. It marked a big step towards semantic search.
A practical illustration: someone types “where can I get my car’s MOT done near Croydon on a Saturday”. Pure word matching looks for pages containing most of those words. A meaning-based system recognises a task (find a garage), a service (an MOT test), a place and a time constraint, and can return garages that open on Saturdays even if no page repeats the query word for word.
Hummingbird arrived as voice and mobile search were growing, and queries were becoming longer and more conversational. It drew on the Knowledge Graph, Google’s database of people, places and things launched the year before, to connect queries to real-world entities. Later additions made the engine better at the same job: RankBrain in 2015 for unfamiliar queries, and BERT in 2019 for the way words relate to each other within a sentence.
Why it matters
Hummingbird is history now, but the change it marked still shapes how a UK business should write for search. Since then, a page has not needed to repeat an exact phrase to rank for it. It needs to answer the need behind the phrase. That is why repeating “emergency electrician Manchester” in every paragraph achieves nothing, while a page that covers call-out times, areas served, Part P registration and what an emergency visit typically involves matches what people actually mean.
It also explains why one strong page can rank for hundreds of variations of the same question. Separate pages for “boiler repair”, “boiler repairs” and “repair my boiler” all serve one search intent, and splitting them now only makes your own pages compete with each other.
Common mistakes
- Treating Hummingbird as something to recover from. If traffic fell recently, look at recent core updates, technical changes or tracking problems, not a 2013 change.
- Still building near-duplicate pages for each keyword variant.
- Writing copy around exact-match phrases so it reads awkwardly, which helps neither readers nor Google.
- Overlooking the context words searchers add, such as “for landlords”, “same day” or “without planning permission”, which often reveal what they really need.
How to act on it
Group your keyword research by meaning, not by spelling. Put phrases that share an intent on one page, and make that page answer the full question: what it is, who it suits, where it is available, roughly what it costs and what happens next. Use the natural vocabulary of the topic, including the related terms an expert would mention without thinking, instead of repeating one phrase. Check the results page for each group to see what kind of answer Google already prefers, whether a guide, a service page or a list of local businesses, and match that format.
Building pages around intent rather than individual keywords is the basis of my content SEO work.