LSI keywords are sold as the secret ingredient of content that ranks: a list of “semantically related” terms you sprinkle through a page so Google understands it. Google’s own answer is short. In 2019, as reported by Search Engine Roundtable, John Mueller wrote: “There’s no such thing as LSI keywords — anyone who’s telling you otherwise is mistaken, sorry.” This post explains where the term came from, why it stuck, and what to do with the reasonable instinct behind it.
What LSI actually is
LSI stands for latent semantic indexing, also called latent semantic analysis. It is an older information-retrieval technique for relating terms and documents: it looks at which words tend to appear together across a collection and uses that to estimate which documents are about similar things, even when they do not share exact words.
That is a real technique with a real history. “LSI keywords”, as sold by SEO tools and courses, is something else: a marketing label attached to lists of related words and phrases. Those lists are not produced by latent semantic indexing, and Google has said the concept they are sold on does not exist. The name borrows the credibility of an academic method to sell a keyword list.
The distinction matters because it changes what you think you are doing. If you believe Google runs LSI and rewards the right co-occurring terms, you will treat writing as a matching exercise. If you accept that the label is marketing, you are left with the honest question: does this page cover what a searcher needs?
Why the LSI keywords myth survives
The myth survives because it contains a half-truth. Pages that rank well often do use a wide range of related vocabulary. A good guide to sourdough will mention starters, hydration, proofing and scoring, because you cannot explain sourdough without them. Tools notice that correlation and present the vocabulary as a cause.
The causation runs the other way. The words appear because the page covers the subject properly. Adding the words to a page that does not cover the subject produces a page that mentions hydration once in a list and explains nothing. That is the same counting habit behind keyword density targets, and it fails for the same reason.
It also survives because it is easy to sell. A list of thirty terms with a coverage score looks like progress. Telling a writer to understand the topic better does not fit in a dashboard.
LSI and semantic keywords: the comparison
The table below separates what the LSI label promises from what actually helps a page cover its topic.
| The LSI keywords approach | What works instead | Where to find it |
|---|---|---|
| A list of related words to insert | Subtopics the page must explain | Headings of pages already ranking |
| Synonyms of the main keyword | The questions searchers ask next | People Also Ask and related searches |
| A coverage score to hit | Entities the topic depends on | Knowledge panels and the topic itself |
| Claimed as a Google method | “There’s no such thing as LSI keywords” | Mueller, reported by Search Engine Roundtable |
| One page per keyword variant | One page per intent, covering its variants | Keyword clustering by shared results |
How to find what LSI keywords promised
People searching for how to find LSI keywords usually want one thing: to know what else a page should cover. That is a fair goal, and you can reach it without the label. Start with the results page for your main query. It is the best evidence you have of what Google thinks the topic includes.
Related searches. The suggestions at the bottom of the results show adjacent queries. Some are separate topics that deserve their own page; others are subtopics your page should answer in a section.
People Also Ask. The question boxes show what searchers want to know next. Each one is a candidate heading or FAQ answer. Our guide to People Also Ask covers how to collect and filter them so you do not end up answering every tangent.
Entities. Every topic depends on named things: people, products, places, concepts. A page about a camera model that never mentions its sensor or its maker is missing the entities the topic is built on. Our post on entity SEO explains how search engines connect them.
Competing pages. Read the headings of the pages already ranking. Not to copy them, but to see which subtopics every page covers and which none does. The first set is the baseline; the second is your opening.
Turning coverage into a brief
Once you have the subtopics, questions and entities, group them. Some belong together on one page; others answer a different intent and need a separate one. Keyword clustering by shared results is the most reliable way to decide, because it uses Google’s own grouping rather than your guess about which phrases are similar.
Then write the brief as a list of questions to answer, not words to include. “Explain how hydration changes the crumb” produces a useful section. “Include: hydration (2x)” produces a sentence written for a tool.
An illustrative example: for a page targeting “how to clean a cast iron pan”, a term list might hand you seasoning, rust, oil and soap to include. A coverage brief asks instead: can you use soap, how do you remove rust, which oil works for re-seasoning, and what to do after cooking acidic food. The words still appear, but now each one sits inside an answer someone was looking for, and an editor can check whether each answer is actually there.
A limitation is worth stating. Mueller’s comment tells you what Google does not use. It does not tell you how Google does model relatedness between terms and topics, which it has not published in detail. Coverage is the best practical approach because it is what readers need and it lines up with everything Google has said publicly, not because anyone outside Google can prove the mechanism. It also means a page can cover its topic thoroughly and still lose to stronger competitors on links or intent match. Coverage earns a place in the race; it does not win it alone.
If you already pay for a tool that produces an LSI keywords list, you do not have to throw it away. Treat its output as a prompt for subtopics you may have missed, check each against the results page, and ignore any score attached to it.
Frequently asked questions
Does Google use LSI keywords?
No. John Mueller said there is no such thing as LSI keywords and that anyone saying otherwise is mistaken.
What is latent semantic indexing?
An older information-retrieval technique that relates terms and documents by how words appear together. “LSI keywords” is a marketing label that borrows its name.
What should I use instead of LSI keywords?
Cover the subtopics and questions searchers have. Use related searches, People Also Ask, entities and the headings of ranking pages to find them.
Are related keywords still worth using?
Yes, when they come from explaining the topic properly. Inserting them into a page that does not cover the subject adds nothing.
The takeaway Drop the LSI label, keep the instinct. Find the subtopics, questions and entities from the results page, and brief writers on what to answer rather than which words to insert.
