Latent Semantic Indexing (LSI): Is It A Google Ranking Factor?

Latent semantic indexing (LSI) is a method of indexing and retrieving information that identifies patterns in the relationships between terms and concepts.

A mathematical technique is used with LSI to find semantically related terms within a collection of text (an index) where those relationships would otherwise be hidden (or latent).

And, in that context, this appears to be extremely important for SEO.

Right?

After all, Google is a massive information index, and we’ve been hearing a lot about semantic search and the importance of relevance in the search ranking algorithm.

You’re not alone if you’ve heard about latent semantic indexing in SEO or been advised to use LSI keywords.

Will LSI, on the other hand, help you improve your search rankings? Let’s take a closer look.

The Claim: Latent Semantic Indexing As A Ranking Factor

The claim is straightforward: optimizing web content with LSI keywords helps Google better understand it, and you will be rewarded with higher rankings as a result.

Backlinko defines LSI keywords as follows:

“LSI (Latent Semantic Indexing) Keywords are conceptually related terms that search engines use to deeply understand content on a webpage.”

You can improve Google’s understanding of your content by using contextually related terms. So goes the story.

That resource goes on to make some compelling arguments in favor of LSI keywords:

“LSI keywords are used by Google to understand content at such a deep level.”
“LSI Keywords ARE NOT SYNONYMOUS.” They are, instead, terms that are closely related to your target keyword.”
“Google does not only highlight terms that are identical to what you just searched for” (in search results). They also highlight similar words and phrases. Needless to say, these are the LSI keywords you want to incorporate into your content.”

Does “sprinkling” terms that are closely related to your target keyword help you rank higher in LSI?

The Evidence For LSI As A Ranking Factor

Relevance has been identified as one of the five key factors that Google uses to determine which result is the best answer for any given query.

According to Google’s How Search Works resource:

“To return relevant results for your query, we first need to establish what information you’re looking forーthe intent behind your query.”

Once intent has been established:

“…algorithms analyze the content of webpages to assess whether the page contains information that might be relevant to what you are looking for.”

The “most basic signal” of relevance, according to Google, is that the keywords used in the search query appear on the page. That makes sense – how can Google tell you’re the best answer if you’re not using the keywords the searcher is looking for?

This is where some people believe LSI comes into play.

If using keywords signals relevance, then using the right keywords must be a stronger signal.

There are tools dedicated to assisting you in locating these LSI keywords, and proponents of this strategy recommend using a variety of other keyword research techniques to identify them as well.

The Evidence Against LSI As A Ranking Factor

Google’s John Mueller has been unequivocal on this point:

“…we have no concept of LSI keywords. So that’s something you can completely ignore.”

There is a healthy skepticism in SEO that Google will say things that will lead us astray in order to protect the algorithm’s integrity. So let’s get started.

To begin, it’s critical to understand what LSI is and where it came from.

In the late 1980s, latent semantic structure emerged as a methodology for retrieving textual objects from computer-stored files. As such, it is an early example of an information retrieval (IR) concept available to programmers.

As computer storage capacity increased and the size of electronically available data sets grew, it became more difficult to find exactly what one was looking for in that collection.

In a patent application filed on September 15, 1988, researchers described the problem they were attempting to solve:

“Most systems still require a user or information provider to specify explicit relationships and links between data objects or text objects, making the systems tedious to use or apply to large, heterogeneous computer information files whose content the user may be unfamiliar with.”

Keyword matching was used in IR at the time, but its limitations were obvious long before Google.

Too often, the words used to search for the information were not exact matches for the words used in the indexed information.

This is due to two factors:

Synonymy occurs when a wide range of words are used to describe a single object or idea, resulting in the omission of relevant results.
Polysemy occurs when the different meanings of a single word result in the retrieval of irrelevant results.

These are still issues today, and you can imagine how much of a headache they are for Google.

However, the methodologies and technology used by Google to solve for relevance have long since moved on from LSI.

LSI created a “semantic space” for information retrieval automatically.

According to the patent, LSI treated the unreliability of association data as a statistical issue.

Without getting too technical, these researchers essentially believed that there was a hidden underlying latent semantic structure in word usage data that they could tease out.

This would reveal the latent meaning and allow the system to return more relevant – and only the most relevant – results, even if no exact keyword match exists.

The most important thing to notice about the above illustration of this methodology from the patent application is that it depicts two distinct processes.

The collection or index is first subjected to Latent Semantic Analysis.

Second, the query is examined, and the previously processed index is searched for similarities.

That is the fundamental issue with LSI as a Google search ranking signal.

Google’s index contains hundreds of billions of pages and is constantly growing.

When a user enters a query, Google searches its index in a fraction of a second to find the best answer.

Using the aforementioned methodology in the algorithm would necessitate Google:

Recreate that semantic space across the entire index using LSA.
Examine the query’s semantic meaning.
Find all similarities in the semantic space created by analyzing the entire index between the semantic meaning of the query and documents.
Sort and rank the outcomes.

That’s an exaggeration, but the point is that this isn’t a scalable process.

This would be extremely useful for small information collections. It was useful, for example, in surfacing relevant reports within a company’s computerized archive of technical documentation.

Using a collection of nine documents, the patent application demonstrates how LSI works. That’s exactly what it’s meant to do. In terms of computerized information retrieval, LSI is primitive.

Latent Semantic Indexing As A Ranking Factor: Our Verdict

While the underlying principles of removing noise by determining semantic relevance have undoubtedly influenced developments in search ranking since LSA/LSI was patented, LSI itself is no longer useful in SEO.

Although it has not been completely ruled out, there is no evidence that Google has ever used LSI to rank results. And Google isn’t using LSI or LSI keywords to rank search results today.

Those who advocate for the use of LSI keywords are latching onto a concept they don’t fully grasp in order to explain why the ways in which words are related (or not) are important in SEO.

Relevance and intent are key factors in Google’s search ranking algorithm.

These are two of the major questions they’re attempting to answer in order to provide the best answer for any query.

Synonymy and polysemy remain significant challenges.

Semantics, or our understanding of the various meanings of words and how they relate to one another, is critical in producing more relevant search results.

However, LSI has nothing to do with it.

Need help with our free SEO tools? Try our free Code to Text Ratio Checker, Broken Links Finder, Online Ping Website Tool.