How to Do a Content Audit That Ends in Real Decisions

· · Content Workflow
Summary

To do a content audit, define one goal, scope a section you can finish, and build an inventory from your sitemap, a crawl and Search Console. Add search data and backlinks, write your decision rules first, then assign each URL one verdict: keep, update, merge or remove. Read the top pages by hand, rank actions by return over effort, and ship the first ten updates before anything else.

Editor's desk with annotated site-map printouts and a laptop open on a spreadsheet

A content audit is a spreadsheet with one row per URL and one verdict per row: keep, update, merge or remove. That is the whole job. How to do a content audit well comes down to three things: a scope you can finish, data you trust, and a rule for deciding that you write down before you look at a single page. Everything else is decoration.

Most audits fail at the last step. The team builds a beautiful inventory, colors a few columns, and then nothing happens, because nobody committed to what a "weak" page actually is. This guide is built around that gap.

What should a content audit answer before you open a spreadsheet?

Start with the question the audit has to settle. "Which pages are underperforming?" is too vague to act on. Better questions are narrow: which posts lost traffic since the last core update, which topics do we cover three times without a clear winner, which pages still describe a product as it was two years ago.

The goal decides your columns. If you are chasing cannibalization, you need queries and landing pages. If you are chasing decay, you need traffic by month. If you are cleaning up before a migration, you need backlinks and internal links. Pick one primary goal and at most one secondary. A seven-goal audit is a seven-month audit.

Scope is the other early decision. Auditing a 40-page site and a 4,000-page site are different jobs. For anything above a few hundred URLs, cut the scope by content type (blog only), by folder, or by age. You can run the second pass later. In practice, what we observe is that audits scoped to one section get finished and acted on, while "the whole site" audits stall at the inventory stage.

Hands sorting printed articles into four stacks for keep, update, merge and remove

How do you build the inventory without missing pages?

One row per URL, pulled from more than one source. A single source always lies by omission.

Use three. The XML sitemap tells you what the site claims to have. A crawl (Screaming Frog, Sitebulb, or the crawler built into your SEO suite) tells you what is actually reachable and how it is linked. Google Search Console tells you what Google bothers to show. The union of the three is your real inventory, and the differences are findings in themselves: URLs in the sitemap but never crawled, URLs that get impressions but are not in the sitemap, orphan pages that only exist in the index.

For each URL, capture the fields that feed a decision and nothing more:

Resist the urge to add forty columns. Every extra column is a place for an argument to start later.

A backlink index earns its place here. Before you remove or merge anything, you need to know which URLs carry links, because those are the ones you redirect rather than delete. Ahrefs and Semrush both export referring domains per URL in one click.

Which metrics decide whether a page stays?

Traffic alone is a trap. A page with 12 visits a month can be the one that closes enterprise deals, and a page with 4,000 visits can be attracting readers who will never buy. So judge each page against the goal you set in step one, using a short list of signals.

Read the trend, not the snapshot. Compare the last 90 days with the same period a year earlier, so seasonality does not fool you. A page that fell 60 percent is a different problem from a page that was always small.

Read the search demand behind the page. If the keyword it targets has no remaining volume, no refresh will save it. If the keyword is healthy and you rank on page two, you have an update candidate with real upside.

Read the links. A page with ten referring domains and no traffic is an asset waiting for a better body. A page with none and no traffic is a candidate for removal.

Read the intent match. Open the current top results for the page's main query. If they are all comparison pages and yours is a glossary entry, the page is not weak, it is mismatched. That distinction changes the verdict from "update" to "rewrite or retire".

Google's own guidance on helpful content is a fair sanity check for the qualitative side: Google's documentation on creating helpful, people-first content lists the questions about expertise, originality and purpose worth asking of any page you are unsure about.

How do you turn data into keep, update, merge or remove?

Write the rules before you review pages. This is the part most audits skip, and it is the reason the verdict column ends up reflecting whoever reviewed the page that afternoon.

Here is a rule set that holds up in practice. Adjust the thresholds to your site's size.

Keep. The page meets its goal, traffic is stable or growing, and the information is current. Leave it alone, but add it to your internal linking targets, because strong pages should pass their strength on.

Update. The URL has demand, links or rankings, and the content is stale, thin, or mismatched to current intent. This is usually the highest-return bucket, since the URL already has authority. Define what "update" means in the row: new data, a rewritten intro, a missing section, a better answer in the first 80 words.

Merge. Two or more pages target the same query or the same intent. Choose the winner by authority (links, then clicks), fold the best material from the losers into it, redirect the losing URLs with a 301, and fix internal links so they stop pointing at redirects. Cannibalization shows up in Search Console when you filter by a query and see several of your URLs splitting the impressions.

Remove. Thin, duplicated or obsolete, with no links, no traffic and no strategic role. If a related page exists, redirect there. If not, return a 410 or noindex it. Deleting is not a failure. The craft, here, is first of all knowing when not to publish.

When you are deciding how deep an update has to go, a content editor helps you compare your page with what ranks. Surfer SEO and Clearscope both score coverage against the current top results, which gives you a concrete answer to "is this stale or just average?". Treat the score as a prompt for judgment, not as the verdict.

Laptop showing a blurred content inventory table beside a notebook of checkmarks

Where do manual reviews beat automation?

Scores and thresholds can sort a thousand URLs into rough buckets in an afternoon. They cannot tell you that a post is accurate, that it sounds like nobody in particular, or that it contradicts your pricing page.

So do both. Let the data produce a shortlist, then read the pages that matter: top landing pages, pages with backlinks, pages on money topics. Fifteen minutes each is enough to catch the problems numbers miss. Check the claims against current facts, check that the author and date are visible, and check whether the page contains anything a competitor could not have written.

If the content was produced with AI assistance, hold it to the same standard, and note it. A machine-assisted page is not weak because of how it was made, but at volume, programmatic sections drift into near-duplicates, and audits are where you find them. Cluster pages by title pattern and compare their bodies. If two pages differ only by a swapped noun, merge or remove one.

Skip the sentiment-style quality score some tools offer unless you can explain how it is computed. A number you cannot defend in a meeting will not survive the meeting.

How do you turn the audit into work that actually ships?

The verdict column is not the deliverable. The deliverable is a ranked list of actions with owners and dates.

Rank by expected return over effort. Updates to pages that sit in positions 4 to 15 with healthy demand come first. Merges that fix visible cannibalization come second. Removals come last, because they save the least effort and carry the most risk, so batch them and verify redirects before launch.

Split the work into sprints of a size your team really completes. Ten updates a week for six weeks beats a 60-item backlog nobody opens. Track the outcome per batch: clicks, impressions and average position for the URLs you changed, measured 4 to 8 weeks after each change. If a batch moved nothing, find out why before you start the next one.

Finally, record the decisions where the next person will find them. A tab in the audit sheet with the rules you used, the date, and the data sources saves the next audit a week.

Wall calendar with sticky notes planning recurring content review cycles

How often should you repeat a content audit?

Once a year for a full pass is a reasonable default for most B2B blogs. Publish heavily, and a lighter quarterly pass on the newest 90 days of content catches problems while they are cheap. After a core update, a site migration or a rebrand, run a targeted audit on the affected section rather than waiting for the calendar.

Between audits, keep the system running. Set up a recurring export of clicks by URL, flag pages whose traffic drops past a threshold you choose, and review that list monthly. A 20-minute monthly check does more for a blog than a once-a-year heroic effort.

Automation helps here. If you use a platform that tracks rankings and page-level changes, such as Clearscope's content inventory or an SEO suite's site audit, schedule the alerts instead of relying on memory.

What would we do first on a site we had never seen?

Three days, in this order. Day one: set the goal, fix the scope, pull the three-source inventory. Day two: add search data and links, write the decision rules, and sort the long tail with them. Day three: read the top 30 pages by hand, finalize the verdicts, and produce the ranked action list.

Then ship the first ten updates before touching anything else. An audit earns its keep when pages change, not when the sheet gets longer. EsyBlog generates and refreshes this type of article on demand. See the demo.

Frequently asked questions

What is a content audit?
A content audit is a systematic review of the pages on a site, usually one row per URL in a spreadsheet, where each page gets a verdict: keep, update, merge or remove. It is based on performance data, links and a quality check.
How long does a content audit take?
A focused audit of a few hundred URLs takes about three working days: one for scope and inventory, one for data and rules, one for manual review and the action list. Larger sites should be audited by section.
How often should you run a content audit?
Once a year for a full pass is a sensible default. Sites that publish heavily benefit from a lighter quarterly check on recent content, plus a targeted audit after a core update or migration.
Which tools do you need for a content audit?
A crawler, Google Search Console, a backlink index such as Ahrefs or Semrush, and a spreadsheet are enough. A content editor like Surfer SEO or Clearscope helps judge how deep an update needs to go.
When should you delete a page instead of updating it?
Remove a page when it is thin or obsolete, has no meaningful backlinks, no traffic and no strategic role. Redirect it to a relevant page if one exists, otherwise return a 410 or add noindex.
What is the difference between a content audit and a content inventory?
An inventory lists what exists. An audit evaluates each item against goals and assigns an action. The inventory is the input, the audit is the decision.