Google’s “Medic” update first hit on 1st August 2018, so this might seem like a strange time to be writing a post about it. Well, I’ve just been putting my slides together for my “What’s New In Search” talk at the Superfast South Yorkshire Digital Showcase Conference on Tuesday and given that much of the discussion around Medic in the immediate aftermath seems to have been misguided, with more recent posts homing in on the real changes Google has been making, now seems a sensible moment to recap.
Named the Medic update because, of all the categories of site affected, sites in the medical sector seemed to suffer the greatest fluctuations, initial assessments in the aftermath of 1st August pointed to changes made to the Google Quality Rater Guidelines (QRG) mere weeks before. A notable section of these guidelines had been updated relating to Page Quality and particularly expanding on Expertise, Authority and Trustworthiness (E-A-T). Jennifer Slegg wrote in detail about the changes here. This was seen as particularly likely to affect “Your Money or Your Life” (YMYL) websites, such as those in the medical & personal finance sectors, which correlated with several analyses of the sites that had been hit hardest.
There followed many blog posts and social media chatter about how to improve E-A-T. Valid as many of these were in the context of improving user experience (and conversion rates), there was some consternation in the SEO community (shared by us at Boom) that (a) it was difficult to see how poor E-A-T could explain many of the examples of sites benefiting or dropping and (b) impressive as Google’s technology is, E-A-T is a very subjective thing and beyond obvious signals like review scores, seems very difficult to build into an algorithm.
Will The Real Medic Please Stand Up
Fast forward a couple of months and the likely true cause of the fluctuations that started on 1st August, and have sporadically been reported since, has become clear. Thanks to in-depth analysis by the likes of AJ Kohn, it seems that “Medic” in fact signalled the introduction of new Natural Language Processing (NLP) features in Google’s algorithm. Nothing to do with E-A-T. Rather, Google is attempting to become better at understanding intent and meaning, matching what users want to the best content through a mathematical approach to textual analysis.
AJ does a far better and more in-depth job of explaining how the area of NLP has progressed to using Neural Embeddings than I ever could, so I recommend heading over to read his post (yes, I’m linking twice to the same post, it’s that good!) He also makes a very important point that Medic was unlikely to ever have been about E-A-T just because of the Quality Rater Guideline updates. The QRGs aren’t a way for Google to check how it’s doing; they’re a way to teach it’s Machine Learning algorithms what “good” looks like. Just like every time you complete a CAPTCHA form you’re training a machine.
Clearly, the machine is still learning. Many of the fluctuations we saw in the early days of August made no sense. Why would a shoe size guide be a better result than a list of women’s shoes in size 9 when you search for “womens shoes size 9”? It seems Google’s assessment of intent isn’t always on the money and this is no doubt why there have been various reports of updates since Medic; either the Google engineers are tweaking the dials, or just as likely, the machine is simply learning more over time as it gets more data (including user metrics from its own search results that help it understand whether it got a particular result right).
On Medical Advice
What practical actions can we take to deal with Medic then? Well, to some extent, doing nothing is a valid option, as AJ demonstrates at the start of his post. As the algorithm gets better, search results will continue to change and you may find that your rankings improve. If you already consider your content to be an excellent answer to your target search terms, one would hope that the algorithm will eventually reward that.
However, even if you haven’t been affected by the update, Google is clearly going to continue down this path, so anyone involved in SEO needs to ensure they’re feeding the machine content in the best way possible. We need to understand something about NLP so that our content is fully “understood”, both literally and in its intent.
Again, others have written better and more in-depth on this subject that I can. Indeed, AJ refers to one such post and I will do the same: Justin Briggs’ “On-page SEO for NLP” is an excellent guide and I urge you to read and digest it before you next write content for the web.
If I were to try and distil the general principle here, it would be to write concisely, in a simple and structured way, clearly answering questions users are likely to have on a subject. Cover the additional questions and topics users are likely to have too; this both aids salience and increases the search terms your page is likely to rank for. If you’re looking for ways to expand on your content by including related keywords and topics, Ann Smarty has just written a good round-up of some tools to assist in finding them.
We have also started performing analysis on pages ranking in the top 10 for target keywords, versus our own pages. We’re looking for common words and phrases in title tags, headings and content generally, along with metrics such as word count. This has been around for a while in the form of TF-IDF analysis of page content, but drilling into specific page elements is so far proving potentially useful.
So there we have it. The Medic update, which in reality didn’t particularly target that sector or YMYL sites more generally, is all about what and how you write and how Google’s NLP works with that language. It adds another dimension when writing for SEO, making you think about sentence structure, formatting and language in a new way, as well as the actual content and keywords your users seek. Happy authoring! 🙂