Most sites have a tail of pages that rank on page two, attract impressions and few clicks, or answer a question badly. Improving those is usually cheaper per unit of traffic than writing new ones, because the page already has some standing. The work is worth buying when that inventory exists, and not otherwise.
Start from the search data, not from a score
The pages worth working on are the ones with impressions and poor click-through, or positions just outside the top results, or a mismatch between what the page answers and what people are asking it. That list comes from your own search console data. Any programme that begins with a tool score instead of your own query data is optimising against the tool.
What the work actually is
Answering the question the page is being asked, earlier and more directly. Adding what is missing rather than padding what is there. Fixing titles and descriptions that misdescribe the page. Merging pages that compete with each other. Removing sections written for a keyword nobody searches. Most useful rewrites make the page shorter and more specific, which is the opposite of what word-count tooling advises.
Content scores measure similarity, not quality
Tools that grade a page against the current top results reward looking like them, which produces sameness and is exactly what search guidance describes as content made for ranking rather than for readers. Use them as a checklist for genuine gaps, never as a target. A page that scores well and says nothing new has no reason to outrank the pages it was copied from.
Measure per page, over a sensible window
Record the position, impressions, clicks and conversions for each page before the change, and read them again after enough time to be meaningful, usually six to eight weeks. Change one page at a time where you can. A programme that rewrites two hundred pages in a month cannot tell you which changes worked.
Questions people ask about content optimization services
Optimise existing pages or write new ones?
Optimise first when you have pages with impressions and weak positions; that inventory is cheaper to improve than to create.
How much can optimisation move a page?
A page already near the top can move meaningfully. A page that is nowhere is usually not an optimisation problem, it is a wrong-page problem.
Should underperforming pages be deleted?
Sometimes. Merging several weak pages into one strong one, with redirects, is often better than either keeping or deleting them.
Is adding words the answer?
Rarely. Answering the question earlier and cutting what does not serve it is more often what moves a page.