Keyword Extractor

What counts as a keyword here?

Anything that is repeated. A single word must be at least three letters long, must not be on the built-in English stop-word list (words such as the, with, your, from) and must appear at least twice. A phrase of two or three words must also appear at least twice. Ranking is by raw count, so a term used once never appears, however important it is. Numbers on their own are skipped, although a number can sit inside a phrase.

Why do I get fewer keywords than I asked for?

The menu sets a ceiling of 10, 20 or 30, not a target. Short texts rarely repeat enough words, so a 75-word email-marketing paragraph returns 16 entries even with 30 selected, and "A short note about cats and dogs." produces none, in which case the tool says no word or phrase appears twice. Paste at least a few hundred words for a usable list.

How do the three keyword types differ?

"Single words only" lists repeated words by frequency. "Phrases" lists the repeated three-word phrases first and the repeated two-word phrases after them, so a three-word phrase seen twice outranks a two-word phrase seen ten times. "Single words + phrases" takes up to 5 three-word phrases, up to 10 two-word phrases and all repeated words, then sorts the lot by count, which means a phrase and its own words (for example "email marketing", "email" and "marketing") can all appear in the list.

How are phrases detected?

Two-word phrases are pairs of neighbouring words in which neither word is a stop word and both are at least three characters. Three-word phrases need a non-stop-word at each end, but the middle word can be anything, so "cost of living" qualifies. Punctuation is ignored when pairing, which means a phrase can span the end of one sentence and the start of the next. Hyphenated terms such as state-of-the-art are counted as one word in the single-word list but are split at the hyphens in the phrase list.

Does the "How will you use these?" menu change which keywords are found?

No. The extraction is identical for every option. The menu changes only the explanatory note under the list, and the "social" option also adds a line of hashtags built from the first 15 entries by deleting spaces, such as #emailmarketing. The SEO note reminds you to check that important terms appear in the title, H1 and opening paragraph. It also points out that Google ignores the meta keywords tag, so there is no need to fill it in.

Does it group "panel" and "panels", or find related terms?

No. There is no stemming, lemmatising or synonym matching: "panel" and "panels" are two separate entries, and text with the same topic described in different words splits its counts between them. The tool also gives no keyword density figure, search volume or competition data. It reports only how often words and phrases occur in the text you paste.

What sort of words slip through the stop-word filter?

The list is short and generic, so frequent but empty words such as "good", "make" or "one" are not removed and can reach the list if repeated. Use the ranking to spot topic terms and ignore filler entries by eye. Because the list is English, other languages get no filtering at all, although accented letters are kept inside words.

Can I use it to see which keywords a competitor’s page emphasises?

Partly. Pasting a rival article shows the terms and phrases it repeats, which is a quick way to see its vocabulary. It cannot tell you where those terms sit on the page (title, headings or body), whether they are what the page ranks for, or how much traffic they bring, because it only counts repeats in the pasted text.