For months, we’ve been fed bits and pieces of information about Google’s BERT, a machine learning technique that promises to be the future of language models and SEO. More than focusing on a couple of keywords, BERT better understands the nuances and context of words – but, as we learned last week, this type of model trained on large datasets could be biased.
Well, what if Google has access to a less well-known and less understood algorithm that can comprehend long-form content even better than BERT? According to Search Engine Journal’s Roger Montti, it does: Meet SMITH.
In a stunning summary of a recently published research paper put out by Google, Montti does a deep dive into SMITH and its ability to understand long pieces of content and influence Google’s passage ranking update.
Some of the most compelling takeaways from Montti’s article are:
- There is no indication of whether or not Google is using the SMITH algorithm. Following its trend of staying tight-lipped, Google has only said that SMITH outperforms BERT, not that SMITH is presently in use.
- BERT is by no means perfect. Its imitations include long-form documents, which the model is not well-suited to understand.
- The SMITH algorithm is trained to predict entire blocks of sentences. Apparently, we’re in the year 3021, because SMITH is not only learning the relationship between words but the context of paragraphs and how they relate to one another.
- SMITH doesn’t replace BERT; it supplements it. Where BERT is unable to perform, SMITH can pick up the slack.
There are no if, buts or maybes about it: SMITH really, truly works. Montti says the authors of the paper state with complete confidence that SMITH is the best model Google has found for understanding language.
So, what does this mean? For the moment, there’s not much SEOs can or need to do because Google hasn’t explicitly stated that SMITH is actively in use. But one thing we know for sure is that passage ranking is imminent, and it seems like SMITH is the ideal model for making it work.
What’s more, the research paper doesn’t mention a single shortcoming of SMITH, painting a near-perfect picture of the model. It seems very likely that, if SMITH is not in use already, it will be very soon. We’ll be following the updates closely because it’s looking more and more like SEO could be in for a shake-up yet again.
More SEO News You Can Use
WordPress Enabled Automated Updates to the All in One SEO Plugin Without User Consent: User reports reveal that the plugin turned on automated updates without warning or opt-ins from users. Predictably, the auto update has been met with less than glowing reviews. In a discussion on a private Facebook group, WordPress site owners aired their grievances with the “ethical issue” of automatic updates not being the user’s choice. To make matters worse, the updates coincided with a major upgrade, causing several sites to suffer outages. On the back of several complaints, the publishers of All in One SEO have since apologized and stated their intention to remove the update.
Using Google Search Console’s URL Removal Tool Does Not Remove the URL From the SERPs and Index: Though it’s a common misconception, Google’s John Mueller has reiterated that the URL Removal Tool won’t remove anything from the index. As Search Engine Journal’s Roger Montti says, the word “removal” does connote a sense of permanence, which is likely where the confusion among SEOs stems from. A question posed on Twitter requested an explanation for why “removed” URLs continued to be indexed, and Mueller explained that the tool results in a temporary removal, the effects of which are, unfortunately, not permanent.
Bing Webmaster Tools Has Been Upgraded With a Bunch of Helpful New Metrics: Although Microsoft Bing doesn’t always get enough credit, it’s doggedly continuing on the path of the righteous, adding a number of new metrics – including crawl requests, crawl errors and indexed pages – to the search performance report. While Bing hasn’t officially posted about the update, an SEO consultant alerted Search Engine Land to the changes. The additional metrics may not be directly related to search performance, but having all the data in one place can give SEOs a deeper understanding of their site’s performance in Microsoft Bing search.
New Data Reveals Almost 40 Percent of All Websites Are Powered by WordPress: WordPress is dominating the internet in a big way. W3Techs, a web technology survey company, has revealed that a whopping 39.6 percent of all websites in 2021 use the WordPress content management system (CMS). Of all websites that use a CMS, WordPress has a market share of 64.1 percent. Perhaps even more surprisingly, the exclusive WooCommerce plugin makes WordPress – not Shopify – the most popular eCommerce CMS. Is WordPress the best CMS? It depends on who you ask and what the requirements are, but you can’t argue the fact that WordPress-built sites that are customizable, UX-friendly and SEO-ready.
A Free, Third-Party Tool Makes It Easier Than Ever to Understand Google Ads Data: The Auction Insights Visualizer, developed by PPC management platform Optmyzr, is a free-to-use tool that helps advertisers visualize and interpret Google Ads data better than before. It’s a win for advertisers who want to share information with clients and managers in a way that’s easier to explain and understand. Google Ads currently presents this data as a table, which isn’t quite as easy to interpret as the customizable charts generated by the tool. The Auction Insights Visualizer requires no registration and minimal manual effort, so it’s definitely worth exploring.
Editor’s Note: “SEO News You Can Use” is a weekly blog post posted every Monday morning only on SEOblog.com, rounding up all the top SEO news from around the world. Our goal is to make SEOblog.com a one-stop-shop for everyone looking for SEO news, education and for hiring an SEO expert with our comprehensive SEO agency directory.