How a content audit works
An audit begins with a complete list of addresses, not a memory of what was published. That list usually comes from crawling the site and reading the sitemap, then joining each page to the data that says whether it works: clicks and impressions from Search Console, sessions and conversions from analytics, and the internal links pointing at it.
Each row then gets a verdict rather than a score. Keep, for pages that earn traffic or enquiries and need nothing. Improve, for pages close to working. Rewrite, where the intent is right but the page is not. Merge, where two pages compete for the same query. Redirect, where the address is worth more than the content. Remove, where neither has value. The verdicts are the deliverable; the spreadsheet is only the working.
Why content audits matter
Most sites grow by addition. Pages are published, nobody removes anything, and after a few years a handful of pages do the work while the rest dilute it. Duplicate topics compete with each other, old service pages contradict the current offer, and crawlers spend effort on addresses nobody wants. None of that is visible from the admin screen, because a list of posts looks the same whether the posts earn anything or not.
An audit is also the safest way to start a redesign or a migration. Deciding what deserves to survive before the addresses change is what makes a proper redirect map possible, and it is the difference between protecting the pages that rank and sweeping everything into the homepage.
Common mistakes with content audits
Judging pages on traffic alone. A page with barely any sessions can be the one a salesperson sends to every prospect, or the one a payment provider expects to exist. Support pages, legal pages and proof pages rarely rank and still need to be there.
Using too short a measurement window is the next trap, because a seasonal page looks dead out of season. And an audit that stops at the spreadsheet is wasted effort. The value sits entirely in the changes made afterwards, so an audit with no owner, no dates and no follow-up is an expensive way to produce a file.
How to act on it
Work in passes rather than all at once. Clear the obvious first: test pages left over from a build, empty tag archives, duplicates, expired offers. Then deal with pages competing for one query, since keyword cannibalisation costs you more than a weak page ever will. Only after that start improving, in the order of the traffic each page could plausibly win back.
Record every decision with a date and a reason, so a change in traffic months later can be traced to something specific. For removals, choose deliberately between redirecting to the closest match and letting the address go, which is the judgment covered by content pruning. Then repeat it on a schedule; a site reviewed regularly never accumulates the same mess twice, and the audit becomes maintenance rather than an SEO audit emergency.