Writing Content to Rank on Google in the Age of Machine Learning
26 March 2020
Machine learning has transformed the way search engines analyse and understand the text on web pages, from the early days of basic word counting to algorithms that teach themselves how to better understand what we’re searching for and which page is the best result.
The key to this is Natural Language Processing (NLP), a collection of techniques that interpret and break down human language into shorter pieces to understand the relationships between words and how they work together to create meaning. You can get an insight into how Google does this using its Cloud Natural Language system here: https://cloud.google.com/natural-language/ (scroll down to the Natural Language API demo).
Try it and you’ll see that Google identifies entities (things, people, places, concepts) and their importance in the text (salience), makes associations between words (hops) and labels them in different ways to understand their function in the sentence. The software even analyses whether the sentiment about the entities is positive or negative.
How does this affect writing content for websites? Essentially, the rules have changed. Out goes the old “write more, pad it out, make sure you use the target keyword x number of times” approach.
Instead, we need to ensure our text is easily understood by NLP algorithms; the more complex our sentences and the more we deviate from the main purpose of the text, the more confused the algorithm becomes as to what the page is about. Less certainty means lower rankings.
Here are some practical tips to write for NLP. Firstly, connect questions to answers by writing a one or two-sentence answer that you’d expect to hear from a voice assistant. Secondly, be clear what you’re referring to. “200 degrees Celsius” is clear, “200 degrees” is not.
In the same vein, keep sentence structures as simple as possible. Check these using Google’s tool described above; the more hops there are, the more convoluted your sentence. Clarity also comes by being specific and absolute. NLP has a hard time understanding “it depends” answers; direct and specific answers are likely to be rewarded.
Be careful that meaning cannot be misinterpreted. “The cat was often in the tree, but now it’s dead” could have two meanings. The same is true of multiple sentences: “Vitamin B12 levels are important. On a TV show last week, Ian said it should be watched carefully.” What should be watched carefully?
You can improve salience (i.e. the likelihood your page is about a certain thing) by the inclusion of terms associated with that thing. Look at the “People also search for” and “People also ask” sections of Google search results for ideas about what other information to include in your page.
Google rewards pages that follow the “searcher journey”, because it has lots of data about the things people search for next after each search. If you already have that information on your page, it’s a better destination for the user, meaning higher ranking for your page.
So avoid the waffle, keep it clear and simple and answer as many questions as you can. Happy writing!
Article written by Ian Lockwood.
