Where these keywords actually help
YouTube's search and recommendation system reads your title, description, and tags to work out what a video is about and who to show it to. Pulling the words and phrases that repeat naturally in your own script or transcript is a fast way to surface the terms viewers already associate with your content - which you can then weave into the title, the first two lines of the description, and the tags field, rather than guessing at keywords that don't actually match what's in the video.
How the extraction actually works
Paste in your script, transcript, or draft description and the tool strips out common filler words (stop words like "the," "and," "is") and counts how often remaining single words, two-word phrases, and three-word phrases repeat throughout your text - anything appearing only once is filtered out, since a true repeated theme is a stronger signal than a word that happens to appear once by chance.
Why single words, pairs, and triples all get checked separately
A single repeated word ("cooking") tells you a broad topic; a repeated two-word phrase ("slow cooker") is often closer to an actual search term people type; a three-word phrase ("slow cooker recipes") can be an exact long-tail phrase worth using verbatim in your title or description. Checking all three lengths at once surfaces candidates at every level of specificity.
Using the output well
Don't dump every extracted term into the tags field - pick the handful that best and most naturally describe the video. This tool is built from your own content, not search-volume data, so it tells you what your video is actually about in the language you used, not how many people search for those exact terms; pair it with a real keyword-volume tool if you want that second half of the picture.