If you are anything like me, you may have a patchy relationship with machine learning. I have gone through phases: from pretending it didn’t exist to thinking I was adding value at a party by talking about it, and everything in between. You may wonder what impact it has on your own writing, or you may not be conscious of it at all. In fact, machine learning has changed dramatically in the last five years, and it does more than affect our income; it determines it.
As important as it is, machine learning isn’t always well understood by writers. There is a gulf between machine learning scientists and the general population, including many of us who actually make money online. The media doesn’t help much as a go-between: they are more likely to promote a scary tale than explain what it does. Also, sometimes it just sounds boring.
What a DJ can teach us about machine learning
Like trying to explain gravity, definitions of machine learning don’t help. I like to compare a DJ at a concert with machine learning in Spotify or YouTube Music. Master DJs seem to intuitively know how to read a crowd and change the atmosphere at will. They can do this without really knowing much about their audience, except maybe what genre of music they have come to listen to. The data points they use are their own music collection, experience, and what they can see.
Modern music services also predict what music we might like, but they use many more data points to do it. As well as your music listening history, machine learning algorithms can see what other people with similar music tastes listen to; by the time of day, location, or device. Every song in a music service inventory is a data point, and machine learning can process millions of those faster than a DJ can twirl a dial.
This isn’t to say that machine learning can replace a DJ yet. But machine learning has advanced in a very short time, almost without us noticing. And in the last five years or so, there have been substantial changes in all of the platforms that many of us hope to make money from. I am not sure that these changes are widely understood yet.
Why should writers care about machine learning and SEO?
In a recent article, Tim Denning recommended that writers ignore SEO, saying that “stuffing common keywords” that people search for won’t make you a remarkable writer. When it comes to writing online, Tim does know what he is talking about. With more than 100 million views, he is one of the most popular and well-recognized writers on Medium. But in this case, I want to suggest we “do as he does,” not “do as he says.”
Tim refers to SEO techniques like overusing keywords and other “black hat” techniques that did help low-quality content rank on search engines in the past. The focus on ranking for search engines, as Tim points out, was bad for readers. It wasn’t just true for Google search engine results but also hashtags on Instagram and keywords for Amazon bookstore search results.
In the last five years, there have been sea-changes in machine learning. For all the negative press about tech platforms, all of these companies understand that their existence depends on users. And unless they are delivering a constantly-improving user experience, they know they can become brands of yesteryear. Overnight.
What do a penguin, a panda, and a hummingbird have in common?
Some years ago, Google founders personally oversaw a complete overhaul of how its machine learning ranked and surfaced content – naming the separate projects after random animals as tech companies do. Now content that shows “Expertise,” “Authority,” and “Trust” is what improved machine learning algorithms that better understand natural speech and context are looking for. The input of humans supplements machine learning too.
What we use now is not your grandparents’ search engine, so to speak. And it isn’t just Google that is making those changes. Machine learning can see when Instagram hashtags don’t actually relate to the content in a post. The same is true for keywords that authors use to promote their books on Amazon. Medium’s own machine learning is geared for users to discover stories, not just for search.
What Does All This Mean for Us Writers?
For a start, it is great news if you are just getting started. In the bad old days, it was very hard to outrank content that had been chosen by the search engines, even when it didn’t deserve it. It means that you can write great quality content now and have a much better chance of being rewarded, even if it still takes time.
More than anything, machine learning evolution means that we should write for humans first and search engines second, as Neil Patel says. It does make sense to research what users are asking about. But we should focus most of our effort on becoming trusted sources of useful, or even expert information.
Which is what I think Tim’s point actually was. It certainly is what he does. Tim’s writing is surfaced by machine learning because he writes with authority and is an expert in the chosen topics. He focuses on what readers want to know, and he is prolific, writing volumes very week. Algorithms pick him up because he has a lot of data points, and now he is established.
Sometimes when we are trying to get there ourselves, it can seem like a very tall mountain to climb, but it’s a journey made easier when we understand how to get there.