How to Run a Content Audit

Key Takeaways

• Decide the question first. "Which pages should we keep" and "which pages are hurting us" produce completely different audits.
• Four verdicts only: keep, update, merge, remove. More categories means less action.
• Merging is the highest-value and least-used verdict. Several thin pages on one topic compete with each other and none of them wins.
• Remove genuinely means remove or redirect. A page left up because deleting feels wasteful still costs you.
• Audit against traffic and intent together. A page with no traffic that answers a question you need to own is a keep.
• Cap the scope. Fifty pages you act on beats four hundred you catalogue.

Decide the Question First

The most common reason audits stall is that nobody agreed what the audit is for, so every page gets evaluated against a vague sense of quality.

"Which pages are worth keeping?" produces a pruning audit. You are looking for thin, outdated and duplicated material and the output is mostly merges and removals.

"Which pages could rank if we improved them?" produces an upgrade audit. You are looking for pieces with real intent behind them and weak execution, and the output is a list of rewrites.

"Where are we missing coverage?" is a gap analysis rather than an audit, and it is a different exercise with a different input, because you cannot find missing pages by looking at existing ones.

"Which pages are competing with each other?" produces a consolidation audit. This is the one most sites need and the one fewest run.

Pick one. An audit trying to answer all four takes five times as long and produces a document too large to act on.

The Four Verdicts

Every page gets exactly one, and the discipline of only four is what makes the audit finish.

Keep. It works, it is accurate, leave it alone. The most under-used verdict, because an audit creates a bias toward finding something to do. If a page is performing and correct, write keep and move on.

Update. The topic is right and the execution is stale or thin. Facts, prices, screenshots, examples, a weak opening, missing depth. Updating an existing page that has some history is almost always cheaper and more effective than writing a new one on the same subject.

Merge. Two or more pages cover the same ground. One becomes canonical, the others fold into it and redirect. This is the highest-value verdict and it is covered in its own section below.

Remove. It should not exist. Genuinely outdated, off-topic, thin with no intent behind it, or an artifact of an old campaign. Remove means delete and redirect to the nearest relevant page, not unpublish and forget.

Force the choice. No "maybe", no "monitor", no "revisit later". Those are the entries that make audits die, because a document that is 40 percent unresolved never gets actioned.

Merging Is the Verdict People Skip

If your site has been publishing for a few years, several pages almost certainly compete for the same query, and this is worse than it sounds.

Splitting one topic across four pages splits everything. Links, authority, and, on any search or answer surface, retrieval. Four mediocre pages on a topic routinely lose to one strong page, and the four keep each other mediocre by dividing whatever attention the topic gets.

Find them by query, not by title. Titles differ; intent overlaps. Group pages by the question a reader would be asking, and any group with more than one member is a merge candidate.

Pick the canonical page on inbound links and history, not on which is best written. The page other things already point at carries the most value, and prose is much easier to move than accumulated links.

Fold, do not delete. Take the genuinely distinct material from each secondary page into the canonical one, then redirect. If nothing on a secondary page is worth moving, that is itself a finding.

Check inbound links before you merge. A page that anchors a cluster, or that other pieces link to for a specific reason, may be doing structural work that its metrics do not show. Those get spared even when they look weak.

What to Measure, and What to Ignore

Metrics support the verdict; they do not produce it.

Worth pulling: traffic over a meaningful window, entry versus total pageviews, the queries a page actually answers, inbound internal links, and last-updated date. Five columns.

Worth ignoring for most audits: bounce rate, time on page, and anything that requires interpretation before it means something. They generate discussion and rarely change a verdict.

Traffic alone is a bad verdict-maker. A page with almost no traffic that precisely answers a question central to your business is a keep or an update, not a remove. A page with decent traffic and no relevance to anything you sell may still be a keep, because it costs nothing, or a merge, if it is competing with something that matters.

Add one qualitative column: is this still true? Prices, product claims, screenshots, "as of" statements. On any site more than a year old this column produces more urgent work than any metric, because inaccurate pages actively damage trust.

Cap the Scope

The failure mode of audits is size.

Cap at fifty pages, or one cluster, or one quarter's output. A capped audit finishes, produces a to-do list, and gets acted on. A four-hundred-page audit produces a spreadsheet with a proud summary and no changes.

If the site is large, audit the cluster that matters most and repeat quarterly. Sequential small audits beat one enormous one, because the actions actually happen between them.

Time-box the sorting. Two minutes per page. If a page takes longer, you are rewriting it in your head rather than sorting it. Write the verdict and move.

End With a To-Do List

The output of an audit is not the audit document.

Convert verdicts into scheduled work before the session ends. Each update is a task with an owner and a month. Each merge is a task with the canonical target named. Each removal is a task with the redirect target named.

Do the removals and merges first. They are quick, they compound immediately by concentrating value, and they produce the visible progress that keeps the effort alive. Rewrites are slower and can be sequenced over a quarter.

Set the next audit date while you are in it. Audits that are not scheduled do not recur, and a site drifts back into the same state within a year.

Record why you made unusual calls. The page you kept despite bad metrics, the page you merged despite decent traffic. In six months somebody will question it, and the reasoning is what stops the decision being silently reversed.

What Tooling Helps

An audit needs a list of pages, some metrics, and a column for the verdict. A spreadsheet is a complete solution.

Analytics gives traffic and entry data. Search Console gives the queries a page actually answers, which is the most useful single input. A crawler such as Screaming Frog gives you titles, word counts and internal link counts quickly. Any of these is optional; none of them makes the verdict for you.

Storyflow is our product and it is not an SEO or analytics tool. It has no crawling, no analytics integration, no rank tracking, no keyword data and no CMS connection, so pull the data with your analytics tool and Search Console and keep the verdicts in a spreadsheet. The adjacent job is the grouping step: laying pages out spatially to see which ones cluster around the same question is genuinely easier on a canvas than in rows, and that grouping is what surfaces the merges. Paid-only during early access, with the Free plan landing before the end of 2026, and anyone a paid member invites to a board joins free now.

The Bottom Line

An audit is a decision session with data attached, and treating it as a data-collection exercise is why so many produce nothing.

Decide the question before you start, because pruning, upgrading and consolidating are three different audits.

Give every page one of four verdicts and refuse a fifth category, however tempting.

Look hardest for merges. Several thin pages on one topic keep each other weak, and consolidating them is the single highest-return action on most established sites.

Cap the scope at something you will finish, then convert the verdicts into scheduled tasks with owners before the session ends.

Do the removals and merges first, because they are quick and they compound immediately.

An audit that ends in a spreadsheet is a census. An audit that ends in a to-do list is an audit.