The definition of new schemes has increased exponentially since its launch in 2011. Below you can see the schemes defined at the time of launch (http:\/\/web.archive.org\/web\/20110728002346\/https:\/\/schema.org\/docs\/full.html)<\/a> compared to those currently available (https:\/\/schema.org\/docs\/full.html)<\/a>. Today we can choose from more than three times as many schemes as in 2011.<\/p>\n As a direct consequence of the extension of schema.org<\/a> and its massive implementation on more and more websites, Google is able to display the answer directly<\/strong> to more queries through the different featured result<\/a> formats.<\/p>\n If in the second half of this year the protagonist has been the application of featured snippets<\/a> such as FAQPage<\/a> and HowTo,<\/a> for next year we foresee:<\/p>\n Otherwise, we continue with the controversy over whether featured snippets generate more or less CTR than a classic first result. As for so many other controversial questions in SEO, the answer is: it depends.<\/p>\n Indeed, it depends on whether the highlighted result is able to fully satisfy the user’s search intent or not. In the first case, the CTR decreases radically, although it is also worth noting the impact on the branding obtained. In the second, the CTR<\/a> actually reaches figures above 70%.<\/p>\n We will have to continue testing to find out how to gain this “extra ball” in the SERPs<\/a> while taking care of the CTR and making the most of the branding<\/em> effect.<\/p>\n You can see here our presentation on the results highlights<\/a> and their impact on voice search, as well as the video of this presentation<\/a>.<\/p>\n The implementation of Bidirectional Encoder Representations from Transformers (BERT)<\/a> is another step in Google’s journey from Natural Language Processing (NLP)<\/a> to Natural Language Understanding (NLU<\/a>), i.e. a concept closer to natural language understanding as we humans do.<\/p>\n The key to reaching a more accurate understanding of a natural text lies in understanding the meaning of each word in the context of the rest of the words that make up a sentence, rather than processing them in a linear fashion, in the same order in which they appear. We can see how Google identifies the entities of a text and the relationship between them with the demo version of the Natural Language API<\/a>:<\/p>\n As we see above, Google knows that Fernando Macia is an entity corresponding to a person, and that it is very relevant to the meaning of the text (0.73). You also know that Human Level Communications is a company. He also understands books as an intellectual work and ten as a number.<\/p>\n In parsing, Google is able to place the function of each word and its relationship to other words in the context of the whole sentence.<\/p>\n As Google better understands natural language:<\/p>\n All this seems to pave the way for Google to truly become an indispensable assistant, capable of understanding spoken natural language and responding<\/strong> in kind with the most relevant data for every need.<\/p>\n\n Tu marca, visible y encontrable<\/p>\n\n\n\n Est\u00e1s a un paso de dominar tu categor\u00eda online<\/p>\n\n\n\nStart now<\/a>\n<\/div>\n\n\n\n Google has already been showing us the entities it relates to a given concept in Google Images results:<\/p>\n The displayed entities help refine the image search in a similar way that Google Suggest suggestions help complete a text search. However, it is likely that Google will also begin to show relationships between entities as the variety of recognized entities and relationships expands:<\/p>\n The prediction at this point is a higher percentage of SERPs where the knowledge graph<\/a> <\/em>appears and <\/em>Google shows other entities to which it relates the searched concept, <\/strong>increasing what we already see in US results:<\/p>\n The recommendation to work on this concept would focus on making it easier for the searcher to identify these relationships as well as sources of authority to help validate them. That is, not only links, but references, even if they do not have links.<\/p>\n Presenting all of these relationships directly in the result pages<\/a> not only facilitate the refinement of the search for the user, but more importantly, statistically validate the relationships between the different entities.<\/strong> That is to say, it will be the users who will confirm to the search engine which are the most relevant relationships between entities for most people.<\/strong><\/p>\n This year there was a great deal of controversy as to whether user behavior might be gaining weight as a relevant factor in SEO. Juan Gonzalez, from Sistrix, already published in 2016 a very juicy post<\/a> with examples of changes in results whose explanation seems to point to the influence of user behavior and Johannes Baus wondered in another (Is user experience a positioning factor?<\/a>) for this same issue, especially after Gary Illyes debunked similar hypotheses put forward by Rand Fishkin<\/a>.<\/p>\n The truth is that, regardless of the way in which Google is incorporating user behavior into its algorithm, any SEO has been able to verify how changes in the positions of different results occur without the classic on<\/em> and off page <\/em>factors having been affected.<\/p>\n![\"Schema.org](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/schema-scaled.jpg\")
<\/p>\n\n
![\"Kayak](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/kayak_faqpage.jpg\")
![\"CTR](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/ctr-1.jpg\")
2. From NLP to NLU<\/h2>\n
![\"Identifying](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/Natural_language_api_demo.jpg\")
![\"Text](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/natural_language_api_syntax.jpg\")
\n
![\"\"](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/HL20-SEO-1-e1770805082974.png\" "\\"HL20-SEO\\"")
\n<\/section>\n\n3. Entities in all results formats<\/h2>\n
![\"Entities](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/entidades_google_images.jpg\")
![\"Entities](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/resultados_fernando_macia.jpg\")
![\"Results](\"https:\/\/www.humanlevel.com\/wp-content\/uploads\/steve_jobs_serp.jpg\")
4. Increased weight of user behavior<\/h2>\n