Google turned fifteen this week, and at its anniversary event it announced, with little emphasis, that it has replaced the core of its search algorithm. The new algorithm is called Hummingbird. It has reportedly been in operation for about a month, and very few people outside Google noticed.
This is notable. Panda and Penguin, the major updates of the last two and a half years, caused sudden ranking losses for many sites and generated extensive discussion in SEO forums. Google states that Hummingbird affects around ninety percent of searches, yet the change went largely unnoticed.
The same week, Google confirmed that it is moving all searches to secure search, which means that keyword data in Analytics will appear as "(not provided)" for nearly all sites. These are two major changes within a few days. In our assessment, the less visible one is the more significant.
What does Google Hummingbird change?
Until now, Google could be described simply as follows: it examines the words in a search and finds pages that contain those words. The actual process is far more sophisticated, but this has been the core principle. As a result, SEO practitioners wrote pages that repeated the words: "Bali dive course. Our Bali dive course is the best Bali dive course in Bali."
According to what Google has disclosed so far, Hummingbird attempts to understand the entire query as a question. Rather than processing "dive", "course" and "Bali" separately, it seeks to identify what the person is actually asking, which words carry the meaning and what the question is about.
In our view, the reason is the growth of phones and voice search. People increasingly speak to their phones, using Siri and Google's own voice search, and spoken queries do not follow search-engine syntax. They take the form of complete questions: "Where can I learn to dive in Bali if I've never done it before?" A system designed to match keywords has difficulty with such queries; a system designed to understand questions does not.
How should content be written for conversational search?
For anyone writing web content, the practical change is clear: build pages around a question a real person has, rather than around a keyword.
Most dive centre Open Water pages we review contain the course name, a price and a promotional paragraph. This is sufficient for someone who already knows what they want. A person who has never dived, however, has many questions. Do I need to be a strong swimmer? How many days does it take? What if I panic? Is it safe for my ears?
A page that answers these questions clearly, in the order people ask them, aligns with the way Hummingbird appears to interpret searches. It also converts more visitors into customers, which has always been the objective.
Why is natural language more effective than keyword repetition?
This change is welcome, since keyword-heavy copy is difficult to write and unpleasant to read.
It is now more effective to write the way your customers speak. Use the words they use, including synonyms, as people naturally do. A diver might say "course", "certification", "getting certified" or "learning to dive". These terms mean roughly the same thing, and Google is becoming better at recognizing this.
We recommend a simple test: read the page aloud. If it sounds like something you would say to a customer at the counter, it is likely appropriate. If it reads as text written for a machine, it should be rewritten.
What does this mean for bilingual businesses?
In Canada, many businesses operate in both French and English, and people do not phrase the same question in the same way in both languages. Their searches differ as well. A francophone looking for accommodation may type a query that is not a translation of what an anglophone would type. A French page translated word for word from the English page therefore misses the questions French speakers actually ask. This was already true before Hummingbird, and it is even more relevant now that Google interprets the question and not only the words. Each version should be written for its own reader. This is the main reason to choose genuine bilingual copywriting over rapid translation.
Can FAQ pages rank for long searches?
Yes. We expect the FAQ page to become considerably more valuable. This does not apply to generic FAQs listing five questions no one has asked, but to FAQs built from the questions customers actually ask by email, by phone and at the counter.
Long, specific questions face less competition. A small dive centre will not outrank major travel sites for "Bali diving". For a precise question about a specific course, asked by a particular type of beginner, a clear answer on a small site can appear near the top of the results. If Hummingbird performs as Google describes, it should improve the matching of these long questions with the right answers.
We recommend adopting the following practices now:
- Collect real questions for a month, recording them in the customer's own words.
- Answer each in a short paragraph that begins with the answer, followed by the explanation.
- If a question requires more than a paragraph, give it a dedicated page and link to it.
- Update the answers when circumstances change. An FAQ showing last season's prices is more harmful than having no FAQ.
What direction is search taking?
With keyword data disappearing and the algorithm interpreting questions, the practice of selecting three keywords and repeating them is losing its effectiveness. The trend we observe points the other way: search engines are steadily improving their understanding of everyday language, spoken or typed, and rewarding the sites that answer questions most clearly.
The pace of this change remains uncertain, but its direction is clear. The pages that perform well under this type of search are the same pages that have always served customers best.
Google has not yet published many details. Further information is likely to be added to the Wikipedia page for Google Hummingbird over the coming weeks as specialists analyse the update. In the meantime, we recommend reading your own course pages aloud.