Photo: Bokeh / Tomedia. People Walking on Historic Wall Path.
At some point, explaining a subject on the internet became a job that apparently required a list of words supplied by an SEO tool. Write the article, count the terms, add the missing ones, and keep going until the indicator changes colour.
I understand the appeal. A score is easier to act on than an editor saying the explanation is incomplete. The problem comes when the score becomes the work, and you start inserting words without understanding what they add.
“LSI keywords” is one of the labels that makes this process sound more scientific than the advice often deserves. LSI is a real information-retrieval technique. Using its name doesn’t establish that Google has given your page a list of related words to include.
Context is ordinary language
Take Apple. You can be talking about a technology company or a piece of fruit. The words around it usually make the meaning clear. iPhones and macOS point in one direction. Orchards and recipes point in another.
A useful page naturally contains the detail that belongs to its subject. The difficulty starts when somebody reverses that observation and decides that adding the vocabulary will provide the understanding.
You can put “wattage”, “efficiency” and “grid-tied systems” in a page about solar equipment without explaining how any of them affect a choice. The reader has encountered the terms and still doesn’t know what to do. A tool may be happier with the page. The person trying to understand an inverter may be no better off.
The appropriate use of a term list is as a prompt to investigate a possible gap. Does this concept belong in the explanation? Can you explain it accurately? Would it help the reader make the decision the page exists to support? Sometimes the answer will be yes. Sometimes the tool has found a word that appears elsewhere and has no useful place here.
A content score won’t write it for you
Some content tools compare how often words appear in your page with how often they appear in other pages. That can help you spot a subject you missed. It doesn’t tell you how many times to repeat a word to make an article good.
The competitor pages used in a content tool are also only a sample. Their language reflects what those pages happen to discuss. If several pages repeat a weak explanation, matching their vocabulary can reproduce the weakness quite efficiently.
Google describes several systems for understanding language and relevance, including BERT and neural matching. There is no reason to reduce all of that to a ritual of placing a phrase the required number of times.
Use the research to ask better questions. If the pages all discuss a feature you omitted, find out why it matters. If none of them answer a practical question you know readers have, that may be an opportunity to do more useful work than reproducing their term distribution.
Read the finished paragraph
This is also why I dislike forcing every keyword variation into headings. You can end up with a page that says the same thing under several slightly different labels, while the actual subject remains underexplained.
Read it as a person who wants the answer. If a sentence exists only because a report wanted the phrase included, decide whether it contributes anything. If it doesn’t, remove it. If a useful idea is present but buried in awkward wording, explain the idea properly.
I would rather see a writer understand why a concept belongs in the page than become very good at making a content score go up. That understanding travels to the next task. A target density generally doesn’t.
The next time a tool tells you the article needs seven more uses of a word, read the paragraph before adding them. If you want help working out which suggestions are useful, The Big Book of SEO on Ephemera covers keyword research and writing pages people can actually use.