"LSI keywords" is an SEO term for words and phrases related to your main keyword, like "orchard", "harvest" and "cider" on a page about apples. The name comes from latent semantic indexing, a document retrieval method patented in 1989, but Google doesn't use it. Google's John Mueller put it bluntly in 2019: "There's no such thing as LSI keywords -- anyone who's telling you otherwise is mistaken, sorry."
The idea behind the term still has a useful core: pages that cover a topic well naturally use the related words, subtopics and names a searcher expects. You get there by covering the topic, not by inserting a list of terms. This guide explains the definition, where the myth comes from, what Google actually does, and a checklist for covering a topic properly.
Key takeaways
- The term "LSI keywords" means terms related to your main keyword. It is SEO slang, not a concept Google uses.
- Latent semantic indexing is a 1988 retrieval method (US patent 4,839,853, granted 1989, expired 2008), designed for document collections far smaller than the web.
- John Mueller of Google, July 30, 2019, said "There's no such thing as LSI keywords".
- Google understands meaning with AI systems such as RankBrain, neural matching and BERT, so it doesn't need a list of co-occurring words on your page.
- What works is covering the subtopics, questions and entities searchers expect. Find them in the search results and top-ranking pages, not in "LSI keyword generators".
What are LSI keywords?
LSI keywords are words and phrases that are conceptually related to a page's main keyword and tend to appear alongside it. For a page about "cold brew coffee", people would call "coarse grind", "steeping time", "concentrate" and "French press" LSI keywords.
They are not synonyms. A synonym replaces your keyword ("cold-brewed coffee"). A related term is something a thorough page on the topic would mention anyway. The SEO advice attached to the term is to include these words so search engines understand your topic.
Here is how the term compares with similar ones:
| Term | What it means | Example for "cold brew coffee" |
|---|---|---|
| LSI keywords | SEO slang for related terms; based on a misunderstanding of latent semantic indexing | "steeping time", "concentrate" |
| Semantically related terms | The accurate name for the same idea: words and concepts connected to the topic | "coarse grind", "acidity", "ratio" |
| Synonyms and variants | Different words for the same query | "cold-brewed coffee", "cold brew" |
| Long-tail keywords | Longer, specific searches you might target with a page | "cold brew ratio for 1 liter" |
| Entities | Specific things Google can identify: products, places, people | "Toddy", "Hario Mizudashi" |
Where the term comes from: latent semantic indexing
Latent semantic indexing (LSI), also called latent semantic analysis, is a method for finding documents by meaning rather than exact word matches. Researchers at Bell Communications Research filed the patent in September 1988, and it was granted in June 1989 as US patent 4,839,853, "Computer information retrieval using latent semantic structure". Its inventors included Scott Deerwester, Susan Dumais and Thomas Landauer. Google lists the patent as expired since 2008.
LSI solved a real problem for 1980s search systems: two documents can be about the same thing without sharing words, and the same word can mean different things ("Java" the island, the coffee or the programming language). LSI builds a large table of which terms appear in which documents, then uses a mathematical technique (singular value decomposition) to group words that tend to occur together into underlying "concepts".
It was designed for fixed document collections, not the web. As Ahrefs notes, the patent calls for re-running the analysis each time the collection changes significantly, which doesn't fit an index that changes every second. At some point SEO writers borrowed the name for "related words", and it stuck.
Does Google use LSI keywords?
No. On July 30, 2019, Google's John Mueller replied on Twitter/X: "There's no such thing as LSI keywords -- anyone who's telling you otherwise is mistaken, sorry."
Google understands language with systems built decades after LSI. Its guide to ranking systems names several:
- RankBrain: "an AI system that helps us understand how words are related to concepts."
- Neural matching: an AI system Google uses "to understand representations of concepts in queries and pages".
- BERT: an AI system that helps Google "understand how combinations of words express different meanings and intent."
- Passage ranking: an AI system that identifies "individual sections or 'passages' of a web page".
These systems understand that a page about cold brew is about coffee whether or not it contains a specific list of words. Adding "related keywords" to a thin page doesn't make it look thorough to them.
Why the LSI keywords idea won't die
Two reasons. First, it describes something real by accident: good pages on a topic do share vocabulary, because they answer the same questions. Tools that compare your page with top-ranking pages will always find terms you haven't used. Second, "LSI keyword generators" are easy to sell. They produce a list of words that sounds technical and gives writers a checkbox.
The risk is in how the list gets used. Forcing terms into sentences produces text that reads like it was written for a machine, and Google's spam policies call out "repeating the same words or phrases so often that it sounds unnatural" as keyword stuffing. A list of related terms is a reminder of subtopics you might have missed, never a quota.
What to do instead: cover the topic
Instead of hunting for "LSI keywords", make sure your page answers everything a searcher for your keyword expects to find. That is what semantic SEO means in practice. Use this checklist before you publish:
Interactive checklist
Topic coverage checklist
0/11
Research the topic, not a word list
0/4Write for coverage
0/4Check before publishing
0/3Your progress is saved in this browser, so you can come back to it later.
Before and after: a sentence-level example
A page built around a list of "LSI keywords" for cold brew tends to read like this:
Our cold brew coffee guide covers cold brew coffee. Use a coffee grind and a coffee filter. Cold brew coffee temperature matters, so use cold water for cold brew coffee.
A page that covers the topic reads like this:
Use a coarse grind, about the size of raw sugar, so the grounds don't cloud the brew. Steep them in cold water for 12 to 24 hours, then strain through a paper filter. Because the water never gets hot, the flavor is different from hot coffee poured over ice.
The second version contains more related terms (coarse grind, steep, strain, flavor) and none of them were inserted. They are there because the explanation needed them.
How related terms fit into a keyword plan
Related terms belong inside a page; keywords decide which pages exist. One page targets one primary keyword and its close variants, then covers the subtopics around it. Our guide on how many keywords to use per page shows the split, and keyword clustering shows how to decide which keywords share a page.
Over many pages, that coverage builds topical authority: Google sees your site as a reliable source on the whole subject rather than one query. If you want to analyze related terms in your own content programmatically, our guide to Python for NLP and semantic SEO shows how.
Outrank builds articles the same way: each one targets a keyword from your plan and is structured around the subtopics and questions searchers ask, with H2/H3 sections, an FAQ and internal links to your related pages.
FAQ
What is an LSI keyword in SEO?
In SEO, an LSI keyword is a word or phrase related to a page's main keyword, such as "steeping time" for "cold brew coffee". The term borrows its name from latent semantic indexing, a 1980s retrieval method, but Google doesn't use that method. The accurate name is semantically related terms.
Does Google use LSI keywords?
No. Google's John Mueller said in July 2019 that "There's no such thing as LSI keywords". Google understands meaning with AI systems such as RankBrain, neural matching and BERT. Covering a topic thoroughly still matters, but not because of LSI.
What did John Mueller say about LSI keywords?
On July 30, 2019, John Mueller of Google replied on Twitter: "There's no such thing as LSI keywords -- anyone who's telling you otherwise is mistaken, sorry."
What is the meaning of LSI?
LSI stands for latent semantic indexing, a method for retrieving documents by meaning rather than exact word matches. It maps which terms appear in which documents and groups words that occur together into concepts. It was patented by Bell Communications Research in 1989.
Are LSI keywords and long-tail keywords the same?
No. A long-tail keyword is a specific, longer search you might target with a page, such as "cold brew ratio for 1 liter". "LSI keywords" refers to related words used within a page to cover the topic. You choose long-tail keywords to decide what to write; related terms appear while you write it.
Do LSI keyword generators work?
They can remind you of subtopics, because they list words that often appear with your keyword. They don't reveal anything Google uses for ranking. You get better input for free from "People also ask", related searches and the headings of the top-ranking pages.
