Did you ever think you’d live in a world where the foxiest job title was “Data Scientist,” people who build “models” all day? Well, Weird Science has come to pass. This is our new and very exciting reality. Knowledge and experiences are wrapped together for a new set of challenges, from data scientist to developer to marketers.
Last year’s news that Google was using a new machine learning tool called RankBrain — used to contribute to its search engine results — caused a kerfuffle in the SEO world, leaving us wondering just what kind of impact it would have.
In general, it seems like the tech industry is feverishly hot on both bots and machine learning (ML) lately. Thousands of data scientist positions are open for hire in Silicon Valley alone. Bots are actually a subset of machine learning called NLP (natural language processing), and the world is buzzing about them.
It is my belief that both AI-bot and ML trends will continue to have substantial yet intriguing impacts on SEO in the years to come.
But the real story isn’t so much about RankBrain, which is just one of several filters Google applies, as what Google is doing with machine learning in general. To understand how this will impact SEO, we need to understand what machine learning really is.
The original definition of Machine Learning is:
A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E.
Machine learning (ML) is not new as a field. It was first proposed in 1959 as “the ability to learn without being explicitly programmed.” The long and the short of it is that ML, while old, is also the new trend du jour. It’s where machines can predict using data models. Data models learn from data. Data is being produced from all sides by the gigabyte… That’s a major reason why Data Scientist will remain the sexiest job title for a while.
Machine learning is more of a statistics-based approach to AI than a lot of people are used to. “Classic AI,” the type that people think about when they think of the horn-rimmed glasses wearing, pocket-protector bearing, MIT or Caltech classic nerd, is about as hip as the outfit they’d be wearing. That was known as “Symbolic” AI, because symbols were used for concepts, and many researchers thought the brain itself was structured this way. Neurobiologists actually joined the party in the 1990s; the field is called Computational Neuroscience – fancy!
ML also applies to people who are accessing data on the internet. You know that any networked GPS game (like Pokemon Go, perhaps) is a geo-specific search, possibly geo-spatial. Geo-spatial means coordinates mapped onto a sphere, with a height component. It’s increasingly common for databases to include geo-search capacities right in their query languages (for example, PostgreSQL stands out here).
OK, so what does ML mean to SEOs? Keywords will become less effective over time. Keywords are the SEOs most influential friend right now, and a lot of less knowledgeable SEOs rely on them exclusively (to the detriment of themselves and their clients).
Geoff Hinton and Jeff Dean at Google have said that algorithm updates Panda and Penguin were based on machine learning systems they designed. The arms race of search algorithms is guaranteed to always put pressure, even the vanguard, to constantly innovate. And as the core ranking systems of search engines get updated with newer systems, they will continue to get exponentially smarter. SEOs themselves should not become machines, but they must harness them.
An Ever-Evolving Search Algorithm
With that in mind, it seems clear that Google’s ultimate aim is to apply machine learning so that its search algorithms will be able to learn and update themselves automatically, and it’s here where the real impact will be felt.
The most noticeable impact, from SEO’s perspective, is that there would likely be much fewer updates such as last year’s “Mobilegeddon” when Google started ranking sites on smartphones and tablets according to their “mobile friendliness.” That major update was implemented by humans, and happened suddenly and all at once. With a machine learning-based search algorithm, it’s likely to evolve more gradually instead of making such sudden changes.
Aside from that, machine learning should be a welcome boon for reputable SEOs, as it will effectively give them a license to continue doing the good work they’ve already started. Google has long stated it requires two things from websites – that they meet its technical requirements, and they also serve as great resources, which means they should contain lots of relevant and knowledgeable content.
Google has always said that websites should first and foremost try to serve up useful content, instead of just keyword-stuffed fluff. It’s likely that machine learning will soon make it even more critical that those requirements are met. Indeed, the evidence suggests that keyword stuffing has already become much less important, as a recent MarketingProfs study points out:
The correlation between keywords and high search rankings has decreased across the board. More and more high-ranking sites are not using the corresponding target keyword in the body, description, or links, the analysis found. Sites are also using keywords less in URLs themselves, with only 6 percent doing so in the 2015 study.
SEOs have long used keyword targeting as a central part of their strategies, but with the introduction of machine learning, keywords simply won’t be necessary anymore. Machine learning means Google will be better be able to understand which are the most authoritative sites for any search phrase, regardless of if they actually use that specific phrase or not, because it will soon be able to understand the actual content on each site.
There’s a reason why Google keeps on pumping out major updates such as Panda and Pigeon. And it may come as a shock to some, but Google doesn’t do it just to frustrate SEOs (that’s just a bonus for them). Rather, it’s all about Google cementing its dominant market share in search, by serving up the best possible results for customers, so they don’t need to look elsewhere. Every single algorithm update has been done with this goal in mind, and the introduction of machine learning has been done for the same reason. Indeed, every search is individualized now, so that when you see results, they are unique to you, but you don’t know why.
As such, maybe all SEOs need to worry about is ensuring their websites are packed with high-quality content that users want to reach and share with their friends. No matter what role machine learning comes to play in Google’s search rankings, that will never change.
What is Deep Learning?
Deep learning is a mix between a buzzword and a concept. The learning part is straight out of AI research and refers to Neural nets. These are data structures that mimic physical neurons in order to simulate brain-like functioning. Recent computing advanced such as cheaper Graphic Processing Units (GPUs) have enabled stacking of neural nets into layers, or even layers or layers. That’s where the deep comes from. Of course, since the field has been around for a long time, the inside joke is that deep learning is really deep marketing.
We dream of machines coming to their own conclusions, yet we fear it at the same time. The funny thing is that ML is not new, nor is it truly automated. Humans still need to engage heavily in training machine learning projects (for now). Self-directed algorithms are a booming field of research, but humans still have to set the parameters.
In search marketing, we’re constantly looking at how people try to find things they need. Search algorithms have always been and always will be at the core of AI research.
In Orwell’s 1984, history was constantly rewritten so that the victors would always find the “right” information in a SEARCH. Studying your potential searchee means following more than keywords, it’s looking at different places and formats. Recent research data indicates that voice search, longer question-style queries are dramatically on the rise.
All images via DepositPhotos.