The question of whether Google and other search engines comprehend content on our websites in a manner akin to human understanding is straightforward: no, they cannot. At least not at present.
However, it is an entirely different matter to consider whether modern search engines are approaching the replication of the process of reading and comprehending text. The answer to this question is a resounding yes.
In this article, we aim to unveil some intriguing facts and dispel certain myths related to search engines, with a particular emphasis on semantics, or the human-like understanding of text meaning applicable to search engines.
What is semantics?
What are semantic search engines?
What is semantic SEO?
Building on these two explanations, we can define semantic SEO as a specific strategy or concept of SEO focused on the optimisation of web content for user intent and context. It goes beyond keywords to enhance relevance and improve search rankings. Semantic SEO is now a standard method in “traditional” SEO services.
Traditional SEO vs. Semantic SEO: A Comparison
| Aspect | Traditional SEO | Semantic SEO |
|---|---|---|
| Focus | Primarily on keywords and phrases | Emphasizes user intent, context, and relationships. |
| Content | Keyword-focused, shorter content | Context-rich, long-form content |
| Keywords | Exact match keywords, high search volume | Long-tail keywords, related concepts |
| Search Ranking Factors | Relies heavily on keyword density and placement. | Considers context, relevance, and user experience. |
| Results Accuracy | May have limitations in understanding context. | Aims for more accurate and contextually relevant results. |
| Drawbacks | Keyword stuffing, poor user experience, susceptible to algorithm updates | Requires deeper understanding of search intent, more content creation effort |
| Future-proof | Less adaptable to algorithm changes | More resilient to algorithm updates, future-oriented |
| Examples | "5 best laptops under $500" | "What are the best affordable laptops with reviews " |
Google’s (re)volution to semantic search engine
Google initiated its dominance in the search industry primarily because its founders recognised that they could outpace the competition only by delivering the most relevant results for any query, rapidly and on a large scale.
As detailed in a previous article, they devised an efficient ranking algorithm that integrated both links and keywords to furnish the most pertinent search results.
However, their journey didn't end there. One of the significant challenges they faced was comprehending what users genuinely sought as a search result when entering specific queries in the browser.
In essence, they needed to instruct their algorithm to understand the meaning behind the text: queries entered by users and content published by webmasters.
Here are the key events and technologies implemented to achieve this formidable goal:
Freebase (2007): Google acquired Freebase, a knowledge graph database, to enhance semantic understanding by incorporating structured data.
Hummingbird (2013): Google introduced the Hummingbird algorithm, focusing on understanding the intent behind complex search queries rather than just keywords.
RankBrain Machine Learning (2015): Google integrated RankBrain, a machine learning algorithm, to interpret and learn from ambiguous or novel search queries.
BERT (2019): Google introduced BERT (Bidirectional Encoder Representations from Transformers) to improve language understanding by considering the context of words in a sentence.
MUM (2021): Google unveiled MUM (Multitask Unified Model), a model capable of multitasking and understanding information across multiple languages.
BARD, the PaLM 2 language model (2022): Google introduced BARD, the PaLM 2 language model, to further advance natural language processing and semantic understanding.
So, owing to these cutting-edge technologies and revolutionary patents, and drawing upon your prior search history, location, previous interests, search queries, and aggregated data from other users (a crucial element for the upcoming section), Google can discern the meaning of the query you searched for and your intent even if you did not use the exact keywords.
In the majority of instances, Google can identify whether you are searching for a place, object, person, organisation, and moreover, whether you seek to acquire additional knowledge on a topic, compare different products, navigate to a specific website, or purchase a particular product.
Shocking Truth: Google Doesn’t Understand Anything!
Something that had been under suspicion by the SEO community for a long time finally received confirmation during the latest antitrust trial where Google is under suspicion that their 90% market share is a consequence not of their great products, but of disallowed business practices, including paying substantial sums to make their products default.
“We do not understand documents. We fake it”, this is what was literally written in the document Google presented during the antitrust trial.
We do not understand documents. We Fake It.
- Today, our ability to understand documents directly is minimum.
- So we watch how people react to documents and memorize their responses.
Google lacks insight into the content of your page, and they don't seem to care much about it.
Given their limited technical ability to comprehend content (they can grasp some basic elements of a document), they don't invest significant resources in assessing the content itself. Instead, they manipulate results and observe user reactions. If a result proves satisfactory—meaning users click on a page, read it, and stay on the page—Google takes note and rewards the page as a good match for that particular query.
This implies that Google currently focuses on users rather than content when making decisions, relying on user signals.
This is precisely what Google aims to conceal, especially from the SEO community. If this is true, manipulating rankings becomes even easier than with other manipulative techniques like keyword stuffing or link spam.
Here are case studies or experiments that demonstrate Google's extremely limited ability to understand text:
Google Relies Heavily on User Signals
This experiment dates back to 2016 when Rand Fishkin, co-founder of Moz and SparkToro, invited his Twitter followers to visit his blog post using specific keywords. About 200 people followed the request, and within a few hours, his blog post jumped from position 7 to position 1.
Google Understands Only Basic Things
This case study is even more remarkable.
Kyle Roof, one of the world's top SEOs, demonstrated that Google doesn't know how to read at all. The machine appears to "understand" things based on certain mathematical calculations. Roof created an article entirely in Latin using a Lorem Ipsum generator, strategically adding terms where Google expected to find them in content it “understands” as quality and relevant. After a few weeks, his entirely non-readable and nonsensical article ranked at position 1 in SERP, Google My Business, and Featured Snippet for the term "Rhinoplasty Plano," surpassing real surgeons in ranking. Following the fame of this case study, Google had no choice but to de-index his website.
Both these examples highlight that Google is far from what they claim to be, and their ability to truly understand text remains minimal. Nevertheless, the importance of semantic, machine learning, and natural language processing will only increase in significance for search in the coming years
All Services
Web Design
Web Development
Ecommerce Solutions
SEO Services
Digital Marketing
Social Media Marketing
Content Marketing
Email Marketing
Branding & Design
Data Measurement and Insights