How to Run a Content Audit: The Keep, Update, Merge, or Remove System That Actually Gets Finished

Most content audits fail because they start as data collection and end as abandonment. If you want to know how to run a content audit that actually changes your organic traffic, the fix is not more columns in a spreadsheet. It is forcing every URL on your site into one of four decisions before you ever open an analytics tool.

I have watched this pattern repeat across the dozens of brands I have run growth for. A founder opens a fresh sheet, exports everything, adds twenty columns of metrics, and by row 60 the whole thing quietly dies. The audit was never really an audit. It was a data hoarding exercise with good intentions.

What a content audit actually is (and why most of them fail)

A content audit is a decision exercise, not a data-collection exercise. You build a complete inventory of every indexable URL on your site, then assign each one a verdict: keep, update, merge, or remove. The data you pull exists only to support those four verdicts, not to fill a spreadsheet for its own sake.

The failure mode is always the same. The team starts with the spreadsheet instead of the decision framework. They pull clicks, impressions, bounce rate, time on page, scroll depth, social shares, readability scores, and fifteen other fields. By the time the sheet is "complete," nobody has the energy left to act on it, and the audit joins the graveyard of half-finished projects in a shared drive somewhere.

The research on this is consistent. Current pruning playbooks explicitly recommend making a yes/no action decision per URL first, then attaching supporting data afterward. The spreadsheet is a record of decisions already made, not a tool for making them. That inversion is the whole game.

Start with every URL, not every blog post

Run a full crawl of your site to capture every indexable URL, not just the posts you remember publishing. Tag pages, category archives, old landing pages, PDFs, and orphaned pages from a campaign two years ago all count. If Google can index it, it belongs in the audit.

Screaming Frog's SEO Spider is the standard tool here, and it is free for up to 500 URLs, which covers most founder-led sites. Set it to crawl, export the full URL list, and you have your content inventory. For larger sites, the paid version removes the cap.

Cross-reference that crawl with your XML sitemap and the Pages report in Google Search Console. The sitemap shows what you think is indexable. Search Console shows what Google has actually indexed. The gap between those two lists is often where the problems hide: pages you forgot about, thin tag archives bloating your index count, or old URLs still live from a site migration.

The reason you audit URLs and not "posts" is that Google ranks URLs. A tag page with 200 words of auto-generated text is competing for crawl budget and diluting your site's overall quality signals just as much as a thin blog post. Treating the audit as a blog-only exercise misses the pages most likely to be dragging you down.

The four data points that actually change decisions

You need exactly four data points per URL to make a reliable keep, update, merge, or remove call. Everything else is decoration that makes the audit harder to finish.

First, pull 12 to 16 months of clicks and impressions from Search Console, broken out by page. Google's own documentation walks through the Performance report export. Sixteen months is better than twelve because it smooths out seasonality. A post that spikes every December looks dead in July if you only pull six months of data, and you will prune something that was about to come back.

Second, pull conversions or assisted conversions from Google Analytics. A page can get modest traffic but punch well above its weight in signups or purchases. Without this, you will delete pages that are quietly making you money.

Third, check backlinks for every URL. Any free backlink checker works for a rough pass. A page with zero traffic but fifteen referring domains is not a removal candidate. It is an update candidate, because those links are passing authority you do not want to throw away.

Fourth, grab word count and last-updated date from the crawl itself. Screaming Frog gives you both. These two fields tell you whether a page is thin (under 300 words with no strategic purpose) or stale (not touched in 18+ months), which feeds directly into the update and remove decisions.

Scoring every page: keep, update, merge, or remove

This is the core of the whole exercise, and it is simpler than most guides make it. Each URL gets exactly one verdict, and the rules for assigning it are mechanical enough that two people scoring the same site independently should agree on 90% of pages.

Keep. The page gets meaningful organic traffic and it converts, or it serves a strategic purpose that traffic alone would not capture. Your pricing page, your best-performing comparison post, a glossary page that earns featured snippets. These pages need no action right now, but they feed into your refresh workflow later.

Update. The page has impressions but low or decaying clicks, or it has backlinks but the content has gone stale. This is the biggest bucket on most sites, and it is where the fastest wins live. A post ranking on page two for a query with real search volume is one good rewrite away from page one. The full refresh workflow for pages marked update covers what to actually change once you have flagged them.

Merge. Two or more pages on your site are competing for the same query, splitting signals and confusing Google about which one to rank. This is cannibalization, and it is more common than most founders realize. The fix is picking a survivor URL, consolidating the best content onto it, and redirecting the rest.

Remove. No traffic, no backlinks, no conversions, no strategic role. The page is dead weight. But this verdict requires the most caution, and I will get into why shortly.

The merge decision: fixing cannibalization without losing rankings

Merging is the step generic guides skip, and it is often where the biggest ranking gains hide. When two pages on your site both target the same query, Google splits its attention between them, and neither ranks as well as a single consolidated page would.

You can spot cannibalization in Search Console by filtering the Performance report to a specific query and checking whether multiple URLs appear. If you see two or three of your own pages swapping positions for the same keyword over time, that is cannibalization in action.

Picking the survivor URL comes down to three signals: which page has the most backlinks, which has the highest current position, and which has the most complete content. Usually one URL wins on at least two of those three. That is your survivor.

Consolidate the best content from the weaker pages onto the survivor. Then set up 301 redirects from the merged URLs to the survivor. Google's own documentation on consolidating duplicate URLs confirms that 301 redirects pass link equity to the target page, which means the survivor inherits whatever authority the merged pages had.

The mistake to avoid is merging pages that target genuinely different intents. Two posts about "email outreach" are not automatically duplicates. One might be a template roundup and the other a strategy guide. If the search intent behind each is distinct, they should stay separate and you should differentiate them more clearly instead.

When deleting content helps, and when it backfires

Content pruning works. The case studies are hard to argue with. Seer Interactive documented a 23% year-over-year organic traffic increase from a pruning engagement. Another case study reported that after pruning, organic transactions rose 4.65% and organic revenue jumped 32.12%. These are not outliers. The mechanism is straightforward: removing thin, low-quality pages concentrates crawl budget and link equity on the pages that deserve it.

But pruning by traffic alone is how you destroy value you cannot see. A page with zero clicks in Search Console might have twenty backlinks pointing at it. Delete it without a redirect and those links hit a 404, and the authority they were passing evaporates. Another page might be earning citations in AI-generated answers without producing measurable referral traffic yet. You will not see that in GA4.

Before you delete anything, run through a short checklist. Check for backlinks first. Check the impressions trend over 16 months, not just clicks. Check whether the page serves any navigational or strategic purpose, like an old campaign page that partners still link to. If a page has backlinks but genuinely no future value, 301 redirect it to the closest relevant page rather than letting it 404.

For pages with no links and no strategic role, you have three options. A 410 status code tells Google the page is permanently gone, which clears it from the index faster than a 404. A 301 redirect works when there is a relevant page to send the equity to. Noindex keeps the page live for users but removes it from search, which is the right call for pages like internal resources or thank-you pages that serve a function but should not rank.

Turn the audit into a work queue, not a document

An audit that ends with a spreadsheet is a failed audit. The output should be a prioritized work queue with owners and dates, because the value is in the execution, not the analysis.

Prioritize by three factors: business value, decay severity, and effort required. A page with high impressions, strong backlinks, and obvious decay is a high-priority update because the potential gain is large and the work is mostly rewriting. A page with zero traffic and zero links is a low-priority removal because the upside is small and the risk is low.

Batch the work in a specific order. Updates first, because they produce the fastest ranking improvements. Merges second, because they require more careful execution with redirects and content consolidation. Removals last, because they carry the most risk if you get them wrong. Practitioner guidance consistently recommends executing in batches and then re-checking Search Console after about four weeks for unexpected impression drops.

Assign each batch to a specific person with a deadline. "Update these 12 posts by March 15" is actionable. "We should update some posts" is not. The audit becomes your editorial calendar for the next quarter, which is exactly what it should be.

How often to audit, and what happens to the pages you keep

For sites publishing new content regularly, run a full content audit quarterly. For smaller sites publishing a few posts a month, twice a year is enough. The cadence matters less than the consistency. An audit you actually finish twice a year beats an ambitious quarterly process that dies after one round.

The pages marked "keep" are not done. They feed into two ongoing workflows. The first is a refresh pass, where you revisit your strongest pages every six months to update statistics, add new sections, and tighten the content against current search intent. The second is running an internal linking audit, which ensures your best pages are passing authority to each other in a deliberate structure rather than a random one.

There is also a newer dimension worth naming. An audit tells you what Google sees: which pages rank, which earn clicks, which carry authority. But the pages you keep and refresh also need to be the ones AI assistants cite when buyers ask questions in your category. Google rankings and AI visibility are related but not identical, and Google rankings no longer guarantee AI visibility. When we built the citation tracking in AnswerRank, the pattern that surprised us was how often a brand's top-ranking page was not the page ChatGPT or Perplexity actually referenced. The cited page was usually the one with clearer structure, more specific claims, and stronger third-party corroboration.

That does not change how you run the audit itself. It changes what you do with the pages you decide to keep. After the refresh pass, it is worth checking whether those pages show up when someone asks ChatGPT or Perplexity the questions your buyers actually ask. AnswerRank's free GPT SEO Checker runs a quick pass on exactly that.

Running a content audit this way, with four verdicts and four data points, is how you actually finish one. The spreadsheet is a byproduct, not the deliverable. If you have been putting off a website content audit because the last one collapsed under its own weight, start with the crawl, pull the four data points, and force every URL into a decision before you open a second tab. Your organic traffic will tell you within a quarter whether it worked.