How to Run a Content Audit in One Afternoon

Most content audits die in the spreadsheet. Someone adds thirty columns, the crawl takes a week and the findings arrive after the team has moved on. A useful one works the other way round: rules agreed before the data opens, a single decision per page, and ten pages to fix this month.

The plan suits a B2B blog or resource centre of roughly two hundred pages or fewer, and expects you to read Google Search Console and export from your CMS. It leaves out deep technical SEO, which needs its own review. You need no paid tool.

Before hour one: agree the rules

Write four decisions on a card: keep, update, merge, remove. Then write one sentence for each saying when it applies. Keep a page that is accurate, useful and still earns traffic or links. Update one whose topic is right but whose facts or opening are weak. Merge one that another page answers better. Remove one with no purpose, traffic or links that cannot be rescued. Agreeing this first stops the afternoon dissolving into debates about individual posts.

Hour one: build the inventory

Pull every URL from the CMS or sitemap and attach a handful of facts to each: when it was published, when it last changed, its clicks and impressions, and a one-line purpose. In the Search Console Performance report, group by the Pages dimension and set the range to twelve months so that seasonal pages are not misjudged. In Google's definitions, a click is a user clicking your site from Search results and an impression is your site appearing there, so the pair says more together than either does alone.

Read the columns in combination. A page with no impressions for a year is a candidate to merge or remove. Impressions without clicks suggest a weak title or opening. A page with both may simply need a date refresh. Never judge on one column.

Hour two: score every page against three questions

Google's helpful-content guidance supplies a ready test, asking who created the content, how it was produced and why it exists, and wanting it made primarily for people. Give each page 0, 1 or 2 on each question, and add one yes-or-no column for whether it is still accurate. A first pass can run at about two minutes a page.

A zero on "why" points to merging or removing. A zero on "accurate" points to updating first, because a wrong page costs trust. Speed matters here, so resist the urge to read whole articles.

Hour three: find overlaps and decide

Group the list by topic. When several pages answer one question, pick the strongest to survive, move the useful parts of the rest into it and redirect the old addresses. Google explains that a permanent redirect signals that the target should be the canonical page, and it advises permanent redirects only when you are sure they will not be reverted. Point each old URL straight at the final destination, not at another redirect, as fixing redirect chains explains.

Favour the address with the strongest links and history over the one with the nicest title. A page that still earns links or traffic should never vanish without a redirect, and every redirect belongs in the sheet. Merging also ends internal competition, giving the survivor a better chance of supplying an AI passage.

Hour four: rank the work and pick ten

Rank the updates by business value: pages near the product, pages sales sends to prospects, pages sitting on the second results page. Pick ten and do those first. Pages in decline show up through the method in finding decaying content.

Then give the ten a short pass for AI answers. The first sixty words should answer the page's question, claims should carry visible sources and the page should show an updated date. Google says AI Overviews and AI Mode have no additional requirements beyond being indexed and eligible to appear with a snippet, so most of this is ordinary editing. Since June 2026 Search Console has also offered generative AI performance reports with impressions by page, so you can check which of your ten already appear in AI features. For deeper work, see optimising existing content for AI search.

An afternoon on invented numbers

These numbers are made up to show the pattern. A team exports 160 pages. Forty-one have no impressions in twelve months. Scoring marks 25 of those as having no purpose, so they are candidates to merge or remove. The overlap pass finds four groups totalling 13 pages, each collapsing into one survivor. The remaining pages marked for update number 34, and the team ranks them, choosing the ten closest to the product and the ones sales already shares.

The site drops from 160 to roughly 125 live URLs, with ten pages rewritten, a redirect log and a short list for next quarter. The typical shape holds: an audit removes more than it adds, and effort goes to the few pages that matter.

Decide in a meeting, not in a database

The usual advice is to gather more data before deciding: backlinks, rankings, scroll depth, conversions per page. We would do the opposite. Extra columns delay the decisions that make the audit useful, and for a site of this size four facts and three scores already separate the clear cases from the debatable ones. Spend the saved time in a thirty-minute meeting where each owner confirms or changes their calls, and treat only the debatable pages as worth further data.

The cost is that a few mid-value pages will be judged on thin evidence, and some owners will defend their old posts. The three questions give a neutral standard for that argument. We would revise this on a very large site, where a sample by section first is wiser than a full pass, or where a page's value depends on data not in these columns, such as sales use. Always ask sales before removing a page with no traffic; it may still close deals.

Special cases that change the plan

Seasonal pages. A guide that peaks once a year looks dead in a short window. Use the twelve-month range and compare the same month last year before you decide.

Pages that serve sales. A resource linked from proposals may have no organic traffic at all. Ask sales which pages they send before touching anything with low numbers.

Pages with a long link history. An old article with strong backlinks is rarely a removal candidate even when its content is weak. Update it, or merge it into a stronger page and redirect, so the links keep working.

Large sites. Beyond a few hundred pages, sample by section first. Audit one section end to end, learn how long it takes, and then decide whether the rest justifies a second afternoon or a standing quarterly routine.

What a decision log looks like

Keep a short log beside the sheet: the page, the decision, the reason in one line, the owner and the date it was done. When a colleague later asks why a post disappeared, the answer is one search away, and the log also shows which kinds of pages keep failing. If four of your ten removals were old product announcements, that tells you something about the publishing habit, not only the pages.

The sheet

Columns that earn their place are URL, title, published and updated dates, twelve-month clicks and impressions, purpose, the three scores, an accuracy flag, the decision, an owner and a due date. Hold to one decision per page, one owner per decision and one date per owner.

After the audit

Record the starting point: indexed page count, total impressions and clicks on the ten priority pages. Check again after sixty days, and do not credit the audit with a change you cannot separate from other causes. A small honest test beats a big claim. Annual audits suit most B2B teams, with a quick look at top pages each quarter. The surrounding programme is described in the complete B2B content marketing guide.