In the past, SEO copywriting involved embedding the words that users utilize in the search engine into a page’s content. So, for instance, someone searching for good pizza restaurants in the vicinity of Qualcomm Stadium in San Diego might search for [pizza near qualcomm stadium] so those are the keywords that your copywriter would embed in the content of the page.
That would help the search engine identify your page’s content with that particular keyword. In the early days, these were basically exact-match searches.
But after a while, Google got better and began using synonyms, so if you were trying to rank a page of your car sales site, you might have used car, automobile, auto or vehicle – and Google would still understand when it saw a search query for car and a page that talked about vehicles, that there might be relevance.
Then in September 2013, Google announced its recent, quiet launch of a new algorithm they called Hummingbird. This was important because Google began to analyze queries semantically, trying to understand what the query was really looking for. It was able to go far beyond simple synonyms, even beginning to extract searcher sentiment in some instances.
A query like [car dealer] might sound rather obvious, but some queries are a lot more vague. As more people began to do voice searches via their smartphones and Chrome browser, they even tended to make more complex queries, often asking complete questions, rather than simply using isolated keywords.
This had a dual effect.
On one hand, more complex queries made it more challenging to understand what the searcher really wanted, because they sometimes tended to speak more like they might to another person.
On the other hand, the query now provided the search engine with more information, so it was more likely to be able to accurately determine the searcher’s intent.
The search engine, for its part, wanted to get the most understanding possible of both the query and the page.
Is This Artificial Intelligence?
Hardly! It’s barely a scratch on the surface of true AI. In fact, there’s a lot of disagreement in the scientific community about what actually constitutes “artificial intelligence.”
AI is almost certainly a goal for Google, but I think there’s still a long way to go before we can say we have undeniably created the ability for a machine to think like a human, if that’s even possible.
Things like sentiment and intent are often difficult to determine accurately, even for a human. Computers are simply mathematical calculating machines; while they can project probabilities to the nth decimal place, their precise manner of “thinking” is often in juxtaposition to our own.
As everyone’s favorite Vulcan, Spock, would say, we are “highly illogical.”
What About Sentiment?
We’ve already seen cases in which Google (and others) were able to distinguish between a predominantly negative or positive sentiment expressed in a document.
If you think about it, elementary analyses of cut and dried pro or con statements is fairly straightforward. Coupling the use of synonyms and similar words/phrases with certain modifiers can make it a simple mathematical problem.
But things like sarcasm, irony, and humor confound algorithms (even more than they do some individuals). For an algorithm, even detecting them as such can be challenging – comprehending them is still next to impossible.
For instance, if you write “Brand X isn’t the worst provider of widgets in the market, however.” you’d wisely be concerned about whether the Google algorithms would include “isn’t” as a modifier of “worst provider in the market” (typically, they will, by the way). Obviously, the difference to the brand could be substantial.
The fact that the algorithms do, in fact, often make that distinction, points to some progress in their development of semantic ability. But semantic mustn’t be conflated with artificial intelligence – the two are quite different.
Is AI Even Possible?
Now there’s a question that’s sure to stir up some heated discussion! Some remarkable scientists insist that AI is, indeed, possible, while other, equally notable minds argue adamantly against the possibility.
For our purposes, as digital marketers, I don’t think we need be concerned with the precise definition or the metrics used to answer the question. We’re only interested in how it affects us in our work.
I’m convinced that the baby steps already achieved toward AI are opening new opportunities for our clients. When I can rank a page for search terms that appear nowhere on the document to either users or bots, and are definitely not synonymous with the terms used in the search query, then call it what you will – it opens new doors.
It allows me to avoid overuse of keywords and concentrate more on readability and ease of comprehension. It lets me create content that can be more emotive and descriptive in order to better engage the readers. At the same time, it provides new fodder for the learning algorithms, allowing them to improve with each iteration.
I don’t know if artificial intelligence, in its most constraining definition, is actually possible. All I know is that as long as the baby steps keep coming, those of us that take advantage of the opportunity will be able to provide better results for our clients.