LSI Keywords and What I Use Instead of Them?
One of the myths about SEO is that there is something called LSI Keywords. Supposedly you could add them to a page to make it rank higher for a specific term or phrase.
There are ways to find and generate these terms or phrases.
The truth is that LSI Keywords are a myth. According to Google patents, there are ways to find terms and phrases that you can add to a page to help it rank higher for words or phrases for which you are optimizing.
Those are not and have never been called LSI Keywords. The first gets described in Google’s many patents about phrase-based indexing (over 20 in all.). The second is domain terms described in Google’s context vectors patent.
What are LSI Keywords in SEO
LSI is Short for Latent Semantic Indexing. It is a method of indexing devised by researchers at Bell Labs in the late 1980s. It indexes small static databases by understanding the connections between words in a document corpus. Those researchers patented the process in the late 1980s and provided an example of indexing eight books as a sample database using LSI. The inventors did not mention LSI Keywords, nor did they suggest LSI keywords to optimize a data set for a specific term or phrase. None of the inventors suggested LSI keywords, and those were not in the invention of LSI. I wrote a post in the past called: Does Google Use Latent Semantic Indexing (LSI)?. The answer is quite likely that they do not use LSI. It was invented and patented in 1988’s Computer information retrieval using latent semantic structure.
Some SEO Toolmakers and some SEOs have written about LSI Keywords who offer those as ways of optimizing content for specific terms and phrases by adding those LSI Keywords to content. There is no proof that LSI Keywords can help optimize any content for particular words or phrases, and SEO tool makers point to sources such as the Wikipedia pages on LSI. The SEOs who suggest the use of LSI Keywords point to a wide range of sources that they claim are LSI keywords, but many of the generators of those sources do not call them “LSI Keywords,” nor do they claim that you can add those to text to better optimize that text for specific terms.
Here are some ways to provide LSI Keywords that do not give you terms or phrases that help you optimize for specific words.
Do not use these methods to find additional terms to add to a page to help that page rank higher for your chosen term. None of them have proved helpful to do that, none of them have had anything to do with LSI, and they developed no LSI Keywords in the generation of those terms.
Google Autocomplete – these are predictions based on autocomplete and possibly a searcher’s previous search history – not intended to help optimize text for specific terms.
Bolded Terms in SERPs – When Google returns search results for a query, Google will show searchers proof that the search results are related to the query used them, and Google will do this by bolding terms in the results for those queries. That reassures searchers that results from a search are related and do not provide anything such as “LSI Keywords” to an SEO. Search engines were built for searchers, not SEOs, and bolding helps searchers.
Many keyword planner tools help SEOs identify keywords to optimize content. Many of these are helpful and are worth using. They do not tell you how to create additional keywords that you add to your content to help that content rank more highly for a specific term or phrase.
Query Refinements at the bottoms of search results. Google sometimes offers a set of query refinements at the end of search results that suggest other things that a searcher can search for in addition to the original query terms selected. These query refinements were not get added to a page about the initial query to help the content be added to it to rank higher for that query. Many patents about query refinements did not mention the use of LSI and were never to get used in that manner. Again, there is no proof that these phrases should be used that way.
Where are LSI Keywords From?
While LSI is a patented technology from Bell Labs, the phrase “LSI Keywords” does not appear in that patent, nor is it the subject of any other patents at the USPTO.
The SEO Tools that sell LSI Keywords do not explain how LSI Keywords help a page be optimized for specific terms or do not include any case studies showing how they work. One does tell us that LSI Keywords are more effective than keyword Density, which has been an SEO Myth for years (there is not and has never been a magical percentage of keyword usage for different niches.)
After looking through most of Google’s patents and papers, there are no papers that describe the effectiveness of LSI Keywords. There are papers on Semantic Topic Models, which have nothing to do with LSI Keywords and much more to do with one of my suggestions for an actual substitution for LSI keywords that may work.
Action Items To Follow In Place of LSI Keywords
If you want to optimize a page for a specific term, there are ways of finding words that improve how your page gets indexed and ranks better for terms that you may have optimized. I say this based on studying patents from Google and adding such terms on pages that brought them more targeted traffic.
Phrase-Based indexing means adding complete phrases on pages that rank highly for a specific word or term and frequently co-occur on those pages. An example is a page that ranks for the phrase, “President of the United States,” which may have frequently co-occurred complete phrases such as “oval office,” “secretary of state,” “rose garden,” and others that taken together predict what that page is about. The phrase-based indexing patents got started at Google in 2004, and there are over 20 related patents on the subject, which means that Google has spent a lot of effort on phrase-based indexing. I wrote about them many times, including a post at: Are You Using Google Phrase-Based Indexing?
Another Google patent describes domain terms on pages that use context vectors to understand terms better than pages may rank for. I wrote about those in Google Patents Context Vectors to Improve Search. The post points out that many terms have more than one meaning and use sources such as knowledge bases, like Wikipedia, to find domain terms to understand better which meaning of a term was intended.
One of the examples from the patent os for the word “horse. To an equestrian, a “horse” is an animal. To a carpenter, a horse is a tool. To a gymnast, a horse is a vault of exercise equipment. If you include domain terms such as “saddle,” “stirrups,” and “thoroughbreds” on that page, those words help a search engine understand that the page is about animals or horses that equestrians might write about.
Adding complete phrases that co-occur on pages, which are indexed in a phrase-based inverted index of the web. Or adding domain terms which Google has also indexed to help define the meanings of terms that have more than one meaning, is a way of adding additional phrases that can help a page rank higher in search results without using LSI Keywords.
There is no use of Latent Semantic indexing. These are methods that Google devised and knew about, and they have not been made up by SEO tool makers or SEOs who are growing myths about SEO. They are worth trying out and exploring in the SEO that you do.
Do you use any tools to mine those phrazes/words?
Hi Olesia,
I usually just use Google to look for frequently co-occurring complete phrases that rank for the same term, and I look at sources such as Wikipedia to find knowledge bases to locate domain terms.
I am not an SEO toolmaker, and it may take a while to find those terms and phrases to add to pages, but it is is worth the effort.
When looking for co-occurring “complete” phrases, how many words are we looking for? I know there is no hard and fast rule, but are we looking for closer to 1-3 words generally, or entire phrases like 5-7 etc.
Just looking to see what I should start looking for.
Hi Joe,
I have used around 4-5 per page in the past. I have also added domain terms on some pages in addition to co-occurring complete phrases.
Bill
Thank you.
Amazing article. Google auto suggestions provide queries that are related to the target keyword along with people also searching for sections. I usually don’t hunt for LSI’s but use what Google recommends
Hi Abduttayyab,
Google has developed some amazing things in the search results that it shows us. Auto suggestions are intended to save us from having to type as much, and get to answers quicker by offering some predictions as to what we might be searching for. Auto suggestions aren’t intended to be added to pages that you are optimizing to help your page b rank higher for the keywords you selected for that page – that is not the point of autosuggestions.
Here is a Google support page that provides more details are what auto suggestions are for, and how they are used and how they are created: Manage Google autocomplete predictions.
I wrote about Google auto suggestions in 2017 in the post How Google Predicts Autocomplete Query Suggestions is Updated.
I also wrote about auto-suggestions back in 2005: Can Google Read Your Mind? Processing Predictive Queries.
Again, the purpose is not to provide you with LSI Keywords. These aren’t intended to be terms or phrases that you add to your optimized page to help it rank higher, because it is unlikely that it will. And again, there is no use of LSI in the creation of auto suggestions, so calling them “LSI Keywords” just doesn’t fit at all. I don’t use autosuggestion in that manner.
Hi really nice article is given by you explained very well and proper explanation with image. i got a lot of idea from this post thanks for sharing the post and keep tough with us
Hi rsoft,
Thank you. I’m glad you appreciated this post.
Thank you for the article and Happy Holidays.
When it comes to Phrase-Based indexing and optimizing existing content to rank higher, would you say that adding unused GSC phrases is a useful starting point?
For example. GSC –> Specific URL –> Sort by Most Clicks –> Add top phrases that don’t currently exist on page.
Hi Evan.
I’m guessing that you mean the keywords that GSC says that your pages are being found for, and I admit that I will look at those pages and see where those pages might rank for those phrases, and if they rank fairly high, I might add those phrases to those pages a few more times to try to get them to rank higher for those phrases.
That is not the use of LSI keywords, but is taking action on terms that a page is actually being found for. I like phrase-based indexing and domain terms because they could possibly help me rank better for the terms that I have researched and selected for a page.
Both methods should work. Chances are you already have those terms on your pages anyway, so getting more traffic to them should be ideal if you could get that traffic.
Bill
hi bill
Thanks for writing this wonderful blog
here is “https://support.google.com/websearch/answer/7368877#zippy=%2Cwhere-autocomplete-predictions-come-from”
In one section, he described the words autocomplete as follows
“”In addition to full search predictions, Autocomplete may also predict individual words and phrases that are based on both real searches as well as word patterns found across the web.””
Isn’t the explanation of the word patterns found across the web questionable? Can we increase the page rank with LSI words?
Hi Pouyan,
Autocomplete suggestions have been created for searchers, and not site owners or SEOs. They are part of the user experience behind SERPs (search results), and their purpose is to make it easier for searchers to complete a query they have started to type, or to choose a query that might be related to what they were going to search for.
Autocomplete suggestions are predictions based either on the search history of a searcher or the query logs related to a query topic that has been started to be typed by a searcher. They were not generated with latent semantic indexing, and aren’t necessarily semantically related to what a searcher may have started to type. They were not intended to be taken from a search box and stuffed into a webpage, and there is no proof anywhere that using autocomplete suggestions could help a page rank higher in search results.
I included domain terms and frequently co-occurring complete phrases in this post because the Google patents that those are from explain how they could cause a page to rank higher for a term or phrase. That just isn’t true for autocomplete suggestion terms, and there is nothing anywhere that says that they will cause a page to rank higher if you add those autocomplete suggestion terms to a page. Those are the reasons why I do not use them in that manner.
Bill
Hi Bill,
Thanks for the article, as always, you provide us with deep and scientifically proven content.
I’m using a feature within SEMrush, called “On Page SEO Checker”, it proposes to include in your page certain terms, used by concurrent pages for the same keyword, as an improvement idea.
Is this the same concept you just tackled in this article?
Yunus
Hi Yunus,
That is something that SEMRush Came up with, and you would have to ask them for any details regarding their tool. It is not reflected in Google’s patents ns not related to the terms that I mentioned. I have not looked at the feature that you are talking about and am not familiar with it.
Hi Bill
Thanks for the greate article.
you said “these are predictions based on autocomplete and possibly a searcher’s previous search history”
I am a bit confused. I think “searcher’s previous search history” are common phrase and if I add them to my article it increase my chance to rank for those phrase too. it it true?
my second question:
how find “frequently co-occurring complete” for my article. for example i write about “best gaming laptop 2022”
Hi Bill,
Wish you a good and happy new year!
You say that “I usually just use Google to look for frequently co-occurring complete phrases that rank for the same term”. I really don’t understand how you use Google for that. Could you please offer some more details?
Thank you very much
Hi Marius,
I search for the query term that I want to rank a page for, and I usually look at the top 10 ranking pages in Google for that query term that match the meaning of the query term that I am trying to rank for. I look for complete phrases that appear on those pages and co-occur a number of times. I look at how they are used on those pages, and rewrite my page to include those phrases. For example. If a page is about Baseball stadiums, I might see phrases the appear frequently on those pages such as “pitcher’s mound,” “concession stands,” “outfield,” “bullpen,” and “playing field.”
Hi Ali,
Autocomplete terms are words or phrases that start with the same letter that a searcher may have typed already when they are starting to type in a query. When a searcher performs a search, and it is similar to something they have searched for in the past, Google may return terms that are related to previous searchers they have performed Look at the phrases that sometimes appear in autocomplete suggestions – their purp[ose is to maybe save a searcher from having to completely type out a query, or predict what they might be trying to type. They aren’t always accurate as to what a searcher might be looking for, and don’t necessarily help an SEO to rank a page higher – that is not why they are three. They are for searchers, and not for SEOs. They may repeat my query term, but there is no guarantee that adding them to my page will help it make more sense to a visitor, or help it to rank higher for my query terms.
Hello Bill,
Thank you so much for great article and it’s been more than 2 weeks I’m reading your blogs.
I want to learn the whole concepts of Google Patents, Sematic SEO and Topical Authority.
Can you please refer me to the post where I can learn them step by step? As I didn’t even understand what patents are!
So, when I read your posts, it’s difficult to understand them. This is first post that I was able to understand to the bits. Just because, I recently watched a short course by Dixon Jones (inLinks founder)
Sending prayers all the way for your quick recovery ????
Hi Umar,. I have been writing about patents since 2005 at SEObythesea.com. In the United States, patents were first filed at the United States Patent and Trademark Office, with George Washington, who was the first patent officer, in addition to the first president. Their purpose is to inspire innovation and encourage others to invent. Google has been filing patents to exclude other search engines from using their intellectual property I have written about many of the patents that Google has filed. You Can look at the categories here to read about patents of different types, including semantic patents
Hi Umar,. I am fine now. I had a stroke at the end of last year. I have relearned how to walk. I have written a post specifically about Semantic search. I will post a link to it in a try to this comment.
Hi Umar
Here is a post specifically about Semantic SEO:
https://gofishdigital.com/blog/what-is-semantic-seo/
It is on the blog of the company I work at:. Go Fish Digital.
Hi Umar,
I don’t have one single post about authority or topical authority. I have written a few posts about authority but not one dpecificsy about topical authority. I will post a link to one I wrote on categorical authority in a reply to this comment. .
Hi Umar, here is may post on categorical authority:
https://www.seobythesea.com/2019/06/categorical-quality/
Thank you so much Bill for guiding me on patents. I’m clear now.
Glad to hear that you’re recovered now.
Simply one more thing that I’d love to know. When creating a content strategy, what we can do to create a better topical map to rank for multiple keywords.
Modern SEO says, and in practice too, now we don’t need to target SINGLE keywords. Gone are the days when you just follow 30 points On Page SEO checklist for optimizing content.
Google picks entities, especially when we are doing for competitive niches. How we can make sure that we’re providing enough information to search engines what about page is about. Especially for eCommerce type sites.
Since the content seems great when ready from the visitor’s perspective.
P.S. You cleared big confusion of autocomplete and other things in this post. Thanks again! 🙂
Hi Umar, I work this post as an example of what could be done when yo good on adding entities to pages rather than keywords:
https://www.seobythesea.com/2014/10/came-love-entities/
A well-written article about the user-specific knowledge graphs! Google makes their algorithm better to offer accurate search results for the search phrases by the users. Great to know that this giant is building the knowledge graphs to offer the best results. The example of search query and its explanation are good. Thanks for your great efforts to explain the things in a precise way, keep doing your great job!