Google Hummingbird: what the 2013 update changed and what it means for SEO now

Google Hummingbird is a change to Google's search algorithm that Google announced in September 2013. Its purpose was to match the meaning of a whole query with the meaning of pages on the web, instead of matching individual words. Google said at the announcement that it had already been live for about a month and affected 90 percent of searches worldwide, according to CNBC's report of 26 September 2013. For SEO, Hummingbird is the reason that writing clear, complete answers to the way people really ask questions works better than repeating a keyword.

This page explains what Hummingbird was, how it differs from the penalty updates people often confuse it with, what it changed in practice, and a checklist for writing pages in the way it favoured. The dates in this page come from the 2013 coverage linked in the text.

What Hummingbird was

Before 2013, search matched pages to the words in a query, with many clever adjustments. A long, conversational question such as "what is the closest place to buy a phone near me" often confused it, because individual words like "place" and "buy" pulled in unrelated pages. Google's search chief at the time, Amit Singhal, described Hummingbird as an effort to match the meaning of queries with the meaning of documents on the internet. PBS reported that it was designed to understand higher level concepts and the relationships between them, in contrast with the earlier Caffeine update of 2010 that focused mainly on words.

Google also gave an example of follow up questions. A person who searches for pictures of a famous dog breed and then asks about common health problems expects the second answer to be about that breed. Hummingbird was part of the move to treat a search as a conversation, which matters again now that people type full questions into assistants.

Hummingbird is not a penalty

This is the point most SEO pages get wrong. Panda and Penguin were filters aimed at low quality content and manipulative links. They could remove a site from view, and a site could recover. Hummingbird was a change in how Google understands queries, so there was nothing to be penalised for and nothing to recover from. If a site lost traffic in the autumn of 2013, Hummingbird was not a penalty it could fix. The cure for any loss is the same as today: find the pages that lost visits and see whether they answer the question searchers are asking. Our checklist on why traffic drops sets out the order for that.

	Panda and Penguin	Hummingbird
What it targeted	Poor content and spammy links	How queries are understood
Effect on a site	Could demote it	Changed which pages matched, with no penalty
How you respond	Remove the problem and wait for a refresh	Write pages that answer the question well

What it changed in practice

• Long and conversational queries. Questions in full sentences returned useful results more often. People began to search the way they speak.
• Synonyms and related terms. A page about "cheap flights" could rank for a query using "affordable airfare" without the exact phrase.
• Topics over keywords. A page that covered a subject completely often outranked one that repeated a phrase.
• Entities and relationships. Search began to treat a person, place or company as a thing with attributes, not as a string. See entity based SEO for how that works.
• Voice and mobile. Spoken queries are long and natural, so the update suited them. Our guide on optimising for voice search picks up this thread.

Later systems built on the same direction. Google has described RankBrain, introduced in 2015, and BERT, introduced to search in 2019, as ways to understand language better. Each step moved further from keyword matching.

A checklist for writing in the Hummingbird way

1. Start from the question. Write down how a real customer would phrase it, and make that phrase an H2 or H3 on the page. Our method for finding those questions is in keyword research in five steps.
2. Answer in the first paragraph. Give the direct answer in two or three sentences, then explain.
3. Cover the neighbouring questions. A reader asking "how much does it cost" also asks "how long does it take" and "what can go wrong". Put them on the same page or link to a page that answers each.
4. Use the words people use, and a few of their synonyms, naturally. If a term has a common alternative, use both once where they fit.
5. Name things precisely. Use the full names of products, places and standards so the page says clearly what it is about.
6. Add structured data where it describes the page. It states in code what the page is about, and Google's introduction to structured data explains the formats. The page on what schema markup does covers it in plain terms.
7. Do not stuff. Repeating a keyword does nothing for understanding. Repetition now makes the page worse to read.

The myth of Hummingbird keywords

Articles still sell advice about "Hummingbird keywords" or "Hummingbird optimisation". There is no such list. The advice that follows from the update is old fashioned and works: choose a topic, understand what the searcher wants, and write the best page for it. A related myth is that Hummingbird made keyword research useless. It did not. It changed what you research, from isolated phrases to the questions and subtopics around a need. The page on the LSI keyword claim deals with a related myth in the same way.

Questions about Google Hummingbird

When did Google Hummingbird start?

Google announced it in September 2013 and said it had been live for about a month by then.

Is Hummingbird still a ranking factor?

It is no longer a separate update you can track. Its approach, matching meaning, is part of how Google handles language, together with later systems like RankBrain and BERT.

What is the difference between Hummingbird and Panda?

Panda was a filter on thin and poor quality content. Hummingbird changed how Google interprets queries. One can demote a site and the other cannot.

How do I optimise for Hummingbird?

Write pages that answer real questions fully, in natural language, with clear names for the things you discuss. There is no separate technique.

Did Hummingbird affect long tail keywords?

Yes. Long, conversational queries became easier for Google to answer, which helped pages that covered a subject in depth.

From matching meaning to naming sources

Hummingbird taught the search engine to read a question the way a person means it, and the assistants of today carry that to its end: a buyer types a full sentence and gets a written answer that names a few businesses. The answer is built from the same kind of pages Hummingbird favoured, those that state clearly who you are, what you do and who you do it for. AI Knows Us asks ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews the conversational questions your customers ask and records which businesses appear in the answer, so you can see how well your pages are being understood by the newest kind of reader.