If you maintain a content library, you need to identify pages with a specific problem before assigning a rewrite. Pick pages to revise because something is wrong with them. Out of date, or addressed to the wrong reader. Content Optimizer finds the second one.
Nothing here tells you a page is declining
There is no report watching your published pages and flagging the ones losing ground. Other products sell that. This one does not, so choosing what to fix is on you.
Where the candidates come from
In rough order of how reliably each one finds a real problem.
Pages that are factually out of date. A product you changed, a price you stopped charging, a process that moved on. Highest priority and easiest to fix. They matter more than they used to, because stale pages get cited by other pages and by AI answers, and the old fact outlives the thing it described.
Pages that rank or get cited and read badly. They already have attention, so anything you fix compounds.
Pages on subjects where you now know more. What you wrote with five customers and what you could write with two hundred are not the same article.
Templated pages that were never distinct. Location and service pages built off one template. Covered below.
Check for a topic or audience mismatch
Take a page you suspect. Select the audience or topic category you meant it to serve and submit the text.
Your category comes back nearest and the piece still does its job. It comes back below others and the emphasis has drifted. It does not come back at all and the page is about something else.
That third outcome, on a page you thought was fine, is the reason to run the check. From inside the draft it is almost invisible.
The editor takes 8,192 bytes at a time and refuses longer text instead of cutting it, so a long page gets checked section by section. Better that way anyway. You find out which section drifted, and not one number for the whole page.
Diagnose before you rewrite
| What you find | Fix |
|---|---|
| Written from the supplier's side, not the buyer's | Restructure around what the reader has to decide, not what you sell |
| Accurate and generic | Add what only you can supply. Usually evidence, not words. |
| Covers several subjects adequately | Split it, or cut it back to one |
Row one is the most common and the hardest to see without the check, because the writing is competent and the subject is right.
The templated-page case
Take two pages built from the same template and compare each against its intended category. Nearly identical nearest-category lists mean they are one page with a name swapped. You now have a measurement instead of a suspicion.
What follows is a business decision. Either you have enough real substance for thirty distinct pages, or five honest ones would serve better. Turning thirty templated pages into thirty slightly different templated pages helps nobody.
What not to do
Do not paste in phrases from the suggestion list to move a number. The measurement cannot tell a term that belonged from one you added to raise a score, and the second option makes the writing worse every time.
Do not rewrite a page that works. Reaching its intended category and being factually current is enough. Age on its own is not a reason.
Do not treat the score as the goal. It tells you whether the writing reads as being about what you intended. Several other properties of a page matter more.
A rhythm that survives
Quarterly: list the pages carrying the most commercial weight and check the facts on each. Factual errors are the highest-value fix and take minutes.
Twice a year: run the category check on the pages you depend on, one section at a time. Rewrite what fails. Leave everything else alone.
When scheduling a refresh, record the mismatch or outdated information first. The writer can then revise the page against a defined problem.
See Content Optimizer, checking a draft before publication, or how comparing meaning helps you edit.