Website Content Strategy: The Audit That Tells You What to Delete

Every page on your site is there because somebody once argued for it. A client asked, a campaign needed a landing page by Friday, a competitor already had one. Nobody wrote the opposite memo. Your site is not a record of the decisions you made. It is a record of the decisions nobody was willing to reverse. How many pages on it could you defend out loud right now, without opening the traffic report first?

Your analytics never show this: they only count the pages people reached. It surfaces in the crawl. Google decides how much of a site is worth crawling and indexing, and when its systems doubt the site as a whole, that number drops. John Mueller described this on the Search Off the Record podcast in July 2026, calling scaled crawled-but-not-indexed pages a possible quality signal, with undifferentiated AI content as a trigger. His advice was a site-level audit, not a resubmission. Your forgotten pages feed that signal.

Why Every Website Content Strategy Turns Into an Addition List

The default state of a page is that nobody finds it. Ahrefs looked at roughly 14 billion pages and found that 96.55% get no organic traffic from Google at all, while another 1.94% receive between one and ten visits a month. Only the remaining sliver has measurable search traffic, so a page with zero clicks is not a failure. It behaves the way most pages do.

Adding is cheap and removing is expensive, and that asymmetry is the whole reason the bias exists. A new page costs one line in a plan. Removing one costs an argument with whoever asked for it. Every brief that goes out, whether to your in-house writer or to content creation agencies, asks for something to be produced.

An SEOmator audit of more than 100,000 sites in the first half of 2026 found 41.7% leaving thin pages under 300 words indexable and 57.3% carrying broken internal links, a median of 14 per site. Vendor data, so directional. The conditions a keep, merge or retire review exists to catch are that common.

None of this is new, and I would rather say so than pretend otherwise. Nielsen Norman Group names unpublishing as the fourth phase of content strategy, and publishes the right questions for it: who makes the final call, what still links to the page, what deletion does to search. Gravity Global says it out loud too: do not trash everything old, and do not keep everything just in case. The questions are published. The mechanism is not: how to choose keep, merge or retire for one URL, what evidence each verdict needs, and in what order to act.

That is also the line between this article and our own content strategy template, which walks a team through writing the document itself, stop-doing list included. This one assumes the document exists and handles the case it does not resolve: a URL nobody is arguing for. If you are still assembling the plan, start there.

The Page-Level Questions a Website Content Audit Has to Answer

A website content audit wants three answers about one page before any verdict. Start with an inventory, not an opinion: URL, owner, last edit, clicks and conversions in a fixed window, links in and out.

Who is this page for, and did anyone arrive on it last quarter?

Name the audience in one sentence. If you cannot, that is the finding: a page written for nobody reads that way. Pull organic clicks and conversions over a fixed window, not sessions. One exception: legal, pricing and support pages have a job unrelated to search. Never retire a page whose purpose was never traffic.

Does it answer a question another page already answers?

This test is about the question, not the wording, which is why a spelling-level duplicate check never catches it. Write the single question each page answers, then sort the list. Two identical sentences mean one page is diluting the other's signals. You need a spreadsheet and the willingness to admit two of your articles were written for the same search.

What is it linked from, and where does it link?

Internal links are the part most teams skip, and they change the verdict more often than traffic does. An orphan page is invisible to readers and crawlers alike, so its zero traffic measures your linking, not its quality. Fix the linking first and the page sometimes earns its place. A page other pages still link to cannot simply vanish. That gap is why merge exists.

The Three Outcomes: Keep, Merge, Retire

Three signals decide the verdict over a fixed window: organic clicks, conversions or a defined non-search job, and external links in. Two or more and the page earns its place. One and it probably merges. None, and it retires. When the evidence is thin only because nothing points at the page, fix the linking first.

Our window is 180 days, and it is a rule we set, not a standard Google publishes. Ahrefs explains why a number is defensible at all: only 1.74% of new pages reach the top ten within a year, and of those, 40.82% get there within a month. Widen it if your buying cycle is longer, but write the number down: a page sitting on a review list for three years without a verdict is the actual disease.

Two cautions before you act. Ahrefs also found that 72.9% of top-ten pages are more than three years old, up from 59% in 2017, so an old page that ranks is an asset rather than a candidate. And the case everyone quotes: after CNET deleted thousands of articles in 2023, third-party estimates put its organic traffic up roughly 29% between June and August. Those figures are estimates, not first-party data, correlation is not causation, and Google publicly warned against pruning as a tactic in the same period. Read it as how the industry reads pruning, not as a promise that pruning works.

One distinction saves arguments later: a page that ranks well and converts nothing is not a retire candidate. The ranking is an asset and the offer is the problem, which makes it a rewrite or a merge.

A Sequencing Rule So the Site Does Not Fall Over

Order matters more than any single decision. Get it wrong and the broken links outlive the page.

1. Name one owner for the URL and record the verdict with the date. A decision without a name gets reopened every quarter.
2. If the verdict is merge, move the content into the surviving page and republish that page first. The survivor has to be better, not just longer.
3. Repoint every internal link, from body copy, navigation and templates, to the survivor or the closest match.
4. Set the redirect to the closest equivalent page, never to the homepage. A homepage redirect tells the reader nothing and wastes the signal.
5. Only now remove or unpublish the page. Check that it resolves and leave it for good, because a redirect deleted later becomes a broken link forever.
6. Remove the URL from the sitemap and confirm it drops out of the index.
7. Recheck at six weeks, not at once: the crawl and the reallocation of position take that long to show up.

Two rules sit on top of the sequence. Never merge two pages with different intents just because both are weak, or you end up with one confusing page instead of two clear ones. And a page that still earns external links has dead content but live authority, which makes it a merge until proven otherwise. If you would rather hand the inventory to someone else, a team on a content marketing services contract can build it in a week. The verdict stays yours.

What to Measure Six Weeks Later

Six weeks is long enough for the crawl to catch up. Check four things: clicks and impressions for the retired URL and its redirect target, whether any link now resolves to a 404, where the merge target sits on the queries the dead page collected, and whether the indexed page count moved.

Expect a small, boring result rather than a lift. Mueller's point still applies: if the quality signal was about the site as a whole, removing ten pages does not restore it on its own. What these checks tell you is whether the decision was clean: nothing broke and the equity landed somewhere.

This is also where the argument should end and a report begin. Checking a page's fate against your own search data, rather than against whoever argues loudest, is why we built Allable to read Search Console and the site inventory side by side.

Conclusion

A website content strategy that never says no is a publishing queue with a nicer name. The pages you keep avoiding a decision on set the tone for everything around them. Pick one URL this week, run the three signals, and write the verdict down with your name and the date beside it. Which page is it?