How to do a content audit that ends in decisions

The stock picture of a content audit is a spreadsheet: two hundred rows, four statuses in four colours, a tab called “to review later” that nobody opens again. That picture is an illustration rather than a survey finding, and the list is not wrong. It is an inventory, and an inventory is not what the exercise was for. Nielsen Norman Group draws the line in its own definitions: a content inventory is a list of every piece of digital content you currently have, while an audit is the pass that evaluates the quality of what the inventory lists. Their audit closes by recording a status for each piece (keep, update or remove) and by assigning an owner to the pieces that need updating. One owner for every page, including the ones you keep or retire, is our extension of that. A name beside a decision is what makes it defensible in three months.

What follows is how to do a content audit that produces one, for whoever runs the library rather than whoever buys the software: a head of content, a marketing lead who inherited four years of published pages. No tool is required, which is not the same as saying no work is required.

An audit is a decision record, not a spreadsheet

The sheet is how you hold two hundred pages in one place while you work through them; it stops being useful the moment it becomes the deliverable, because then the audit has become a research phase and a research phase has no end. Rows can always be added and columns made more precise, and each extra one postpones the afternoon when somebody has to say what happens to a page.

A website content audit that ends as a list has produced a list. Decisions can be argued with and lists cannot. Once a page leaves with one fate and one name beside it, a colleague can disagree with the fate — and disagreeing then costs a conversation instead of a rewrite. When the audit ends in a number out of a hundred, there is nothing to disagree with and nothing happens next.

What the audit has to produce: one fate per page, one name per page

Two columns do most of the work. The first holds the fate: keep as it is, update, merge into another page, or retire — the same four used in every article of this cluster. The second holds the name of the person who owns that fate, meaning the person who can be asked about it in three months and will remember the reasoning. A third column for the reason makes the decision reviewable rather than merely recorded.

A page with no name is a page that will not change. In our own audits the statuses and the reasoning are present and the missing piece is the person on the hook for any of it, so the sheet gets filed and the pages stay as they were. That is Senatore’s reading of where audits stall rather than a measured finding: naming the owner is what turns a finding into a task.

The four inputs worth collecting, and the ones that only look thorough

The four inputs that carry an audit

Four of them do the work. What the page is: its URL, its title, and the question it was written to answer. What you already know about it: the clicks and impressions trend you can read for free in Search Console’s performance report, which is the only performance evidence this method needs. What it claims: the one-sentence position the page takes, which is what you are really deciding whether to keep. When it last changed substantively, and by whom: a date and a name, which together tell you whether the page is a maintained asset or a fossil with a recent timestamp.

The inputs that only look thorough

Those measure a page instead of judging it: word count, keyword density, a plugin score out of a hundred, the bounce rate of a page whose job is to be read once and acted on. None of them tells you whether the page is still true, still necessary, or still worth the domain it sits on.

How to judge a page without opening a tool

Read it as the stranger who arrives from a search result, and ask four questions of the text itself. Does it answer the question it was written for, in the first screen? Is its central claim still true? Is anything in it now false — a price, a product name, a person in a role they have left, a rule that has been superseded? Would you send it to a customer who asked about the subject this afternoon?

That is not a homemade standard. Google’s guidance on creating helpful, people-first content is in substance a self-assessment list, and factual accuracy is one of the items on it: for informational pages, the content should be factually accurate. A criterion published in a search engine’s own documentation is easier to defend to a client than one invented in a workshop.

The entitlement question sits underneath it. A page is worth updating while it is still entitled to make its claim in public, which is the argument we set out in why SEO is an editorial decision before it is a technical one. If the claim has been overtaken, no amount of editing rescues the page; it belongs in the merge or retire column instead.

Working the four fates in the right order

Order the work cheapest first: update, then merge, then retire, with keep-as-it-is as the default that has to be justified rather than the option that has to be earned. Updating has the smallest blast radius: the URL is unchanged, the history is intact, and existing readers notice only that the answer got better. That work is set out in how an old article is updated without starting over. Retirement comes last because it is the only fate that can end with the address pointing nowhere, and because the address is harder to rebuild than the text.

GOV.UK’s publishing guidance treats the decision as graded rather than binary: you can either withdraw the content - this keeps the content on the site, but it does not appear in internal search results, or unpublish it, which removes the content from the site completely and lets you set up a redirect to current content. The same page’s standing advice is that guidance of this kind should usually be updated rather than withdrawn. That guidance is written for one publisher’s processes and is not a Google requirement; it is useful because it shows how many settings the decision has. Whether the address stays, points at a survivor or is left to fail is a separate decision from the fate, and what happens to that address is where it gets made.

Where audits quietly fail

They fail at the end, in unremarkable ways. Statuses get assigned and owners do not, so nothing moves. A second approver is added late, and two produce the same outcome as none. Scope creeps to every URL ever published, and the audit is treated as preparation for a strategy that never gets written, so the sheet becomes the strategy by default.

One more failure is worth naming, because it looks like diligence: auditing only the pages that are easy to judge. A pricing page with one obvious defect gets fixed while the twelve pages carrying the site’s actual subject matter stay untouched, because those need an opinion. If the audit closes without a single uncomfortable judgement in it, it has measured the library instead of deciding anything about it. Our own version of the work, and what it delivers, is described on the SEO and content page.

Twenty pages, decided in one sitting

Take twenty pages rather than two hundred, and pick them badly on purpose: the ones you would least like a customer to read, the ones whose traffic has fallen, the ones nobody can explain. Put four columns beside them (fate, owner, reason, date) and finish all twenty in one sitting, even if several fates are provisional; those four columns are the whole content audit template this method needs. Twenty decided pages change a library. Two hundred undecided ones change the formatting of a spreadsheet.

No outcome is guaranteed: a decided library is a better library, and what a search engine does with it next is not ours to promise. Send Senatore the page you cannot decide about. One page tells us more about a library than any inventory of it.

---

Written by Senatore. This is Senatore’s own writing about craft; it contains no client claims and no statistics.

Content audits and page libraries we run for clients · why SEO is an editorial decision first

Back to the blog · More in SEO & content