Group items tagged

“Wouldn’t it be nice if Google understood the meaning of your phrase, rather than just the words that are in the phrase?” said Google’s Eric Schmidt back in March 2009, just before the company announced rolling out their first semantic offerings.This signaled one of the first moves away from “strings to things,” and is perhaps the advent of entity-oriented search implementation by Google.

On the whole, however, much of language can be resolved by mathematical computations around where words live together (the company they keep), and this forms a large part of how search engines are beginning to resolve natural language challenges (including the BERT update).

Google’s team of linguists (Google Pygmalion) working on Google Assistant, for example, in 2016 was made up of around 100 Ph.D. linguists.

By 2019, the Pygmalion team was an army of 200 linguists around the globe

BERT in search is mostly about resolving linguistic ambiguity in natural language. BERT provides text-cohesion which comes from often the small details in a sentence that provides structure and meaning

BERT is not an algorithmic update like Penguin or Panda since BERT does not judge web pages either negatively or positively, but more improves the understanding of human language for Google search.  As a result, Google understands much more about the meaning of content on pages it comes across and also the queries users issue taking word’s full context into consideration.

BERT is about sentences and phrases

We may see this reduction in recall reflected in the number of impressions we see in Google Search Console, particularly for pages with long-form content which might currently be in recall for queries they are not particularly relevant for.

International SEO may benefit dramatically too

Question and answering directly in SERPs will likely continue to get more accurate which could lead to a further reduction in click through to sites.

Can you optimize your SEO for BERT?Probably not.The inner workings of BERT are complex and multi-layered.  So much so, there is now even a field of study called “Bertology” which has been created by the team at Hugging Face.It is highly unlikely any search engineer questioned could explain the reasons why something like BERT would make the decisions it does with regards to rankings (or anything).Furthermore, since BERT can be fine-tuned across parameters and multiple weights then self-learns in an unsupervised feed-forward fashion, in a continual loop, it is considered a black-box algorithm. A form of unexplainable AI.BERT is thought to not always know why it makes decisions itself. How are SEOs then expected to try to “optimize” for it?BERT is designed to understand natural language so keep it natural.We should continue to create compelling, engaging, informative and well-structured content and website architectures in the same way you would write, and build sites, for humans.

In database storage mode, you no longer need to save crawls (as an .seospider file), they will automatically be saved in the database

You can de-select specific page elements from being crawled and stored completely to help save memory.

SERP ownership trackers: are increasingly common. For instance Visably (I have no affiliation but do have a free account, which you could get too) and SERP Sketch (I’ve heard of it but not used it). Rather than just telling you if you’re ranking for a specific keyword, they’ll show you if any of the pages in the top ten results mention your brand. That’s the kind of information you can use to select outreach targets.

Content must be user-centric and anticipate needs and questions along the customer journey, just like Google does with MUM

For SEOs, this means that in the future, when designing the content of audios and videos, they will be able to pay attention to a semantically meaningful design similar to that of text, by using keyword research or TF-IDF analyses

Semantic databases like the Knowledge Graph will also benefit from the additional sources of actionable information about entities for data mining

Tools like Buzzsumo and Text Optimizer can give you a good insight into questions people tend to ask around your topic:

Note that Search Console labels featured snippet positions as #1 (SEO used to call them position 0). So when you see #1 in Google Search Console, there’s nothing to do here. Focus on #2 and lower.

MyBlogU (disclaimer: I am the founder) is a great way to do that. Just post a new project in the " Brainstorm" section and ask members to contribute their thoughts.

1. Aim at answering each question concisely My own observation of answer boxes has led me to think that Google prefers to feature an answer which was given within one paragraph. An older study by AJ Ghergich cites that the average length of a paragraph snippet is 45 words (the maximum is 97 words), so let it be your guideline as to how long each answer should be in order to get featured. This doesn't mean your articles need to be one paragraph long. On the contrary, these days Google seems to give preference to long-form content (also known as " cornerstone content," which is obviously a better way to describe it because it's not just about length) that's broken into logical subsections and features attention-grabbing images.  Even if you don’t believe that cornerstone content receives any special treatment in SERPs, focusing on long articles will help you to cover more related questions within one piece (more on that below). All you need to do is to adjust your blogging style just a bit: Ask the question in your article (that may be a subheading)Immediately follow the question with a one-paragraph answerElaborate further in the article

2. Be factual and organize well Google loves numbers, steps and lists. We've seen this again and again: More often than not, answer boxes will list the actual ingredients, number of steps, time to cook, year and city of birth, etc. Use Google’s guide on writing meta descriptions to get a good idea what kind of summaries and answers they are looking to generate snippets (including featured snippets). Google loves well-structured, factual, and number-driven content. There's no specific markup to structure your content. Google seems to pick up <table>, <ol>, and <ul> well and doesn't need any other pointers. Using H2 and H3 subheadings will make your content easier to understand for both Google and your readers. 3. Make sure one article answers many related questions Google is very good at determining synonymic and closely related questions, so should be you. There's no point in creating a separate page answering each specific question. Creating one solid article addressing many related questions is a much smarter strategy if you aim at getting featured in answer boxes. This leads us to the next tactic: 4. Organize your questions properly To combine many closely related questions in one article, you need to organize your queries properly. This will also help you structure your content well. I have a multi-level keyword organization strategy that can be applied here as well: A generic keyword makes a section or a category of the blogA more specific search query becomes the title of the articleEven more specific queries determine the subheadings of the article and thus define its structureThere will be multiple queries that are so closely related that they will all go under a single subheading For example: Serpstat helps me a lot when it comes to both discovering an article idea and then breaking it into subtopics. Check out its " Questions" section. It will provide hundreds of questions containing your core term and then generate a tag cloud of other popular terms that come up in those questions:

5. Make sure to use eye-grabbing images

How about structured markup? Many people would suggest using Schema.org (simply because it's been a "thing" to recommend adding schema for anything and everything) but the aforementioned Ahrefs study shows that there's no correlation between featured results and structured markup.

Technical Guidelines for Article Links There are two technical requirements that are recommended by Google in order to be featured in Discover feeds, listed below. These recommendations apply only to the ‘Article Link’ content types. These technical guidelines do not apply to YouTube videos or shorts, web stories or Ads.

Images are a major part of the Google Discover experience

RSS Feeds

Follow Feature

quality of its content

Provide content that's timely for current interests, tells a story well, or provides unique insights

The “shelf life” of an article within a Google Discover feed may only be 1 or 2 days.

According to a Search Engine Journal study, 46% of a sample size of Google Discover URLs were news sites and 44% were Ecommerce.

It's important to note that impressions are only counted when a link from your site is scrolled into view.