How to Run a Content Audit: Keep, Improve or Cut

Most of what gets called a content audit is an inventory with a traffic column. Every URL on the site is listed, a figure is pasted next to each row, the file is shared in a meeting, and nothing on the site changes. Six months later the same pages are still there, doing the same nothing.

I have run content audits for Australian businesses for 13 years, and the reason they stall is almost never the research. It is that nobody wrote down what was supposed to happen to each page, or who was going to do it by when.

This is written for the owner or marketer who suspects their site has accumulated pages nobody chose. It covers building the inventory from evidence rather than opinion, scoring each page against the job it was built to do, judging quality against criteria a search engine would recognise, and recording every decision so it gets acted on. The scope is narrow on purpose. This is a content audit, not an SEO audit, and it does not go into writing technique or keyword research.

What a content audit is, and what it is not

The one question an audit has to answer

A content audit is the process of reviewing the pages on a site and deciding, page by page, whether each one should be kept, improved, merged, redirected or removed. That is the entire job. Everything else in the process exists to make those five decisions well.

Every row has to answer one question: is this page still doing a job that matters, and is it doing that job well enough to justify the effort of keeping it? If a page cannot answer that question, the audit is not finished, however tidy the spreadsheet looks.

The audits that fail are the ones with no decisions in them. I have watched a six month project produce a careful inventory, a scoring model with nine columns, and not one merge, redirect or removal. The work was thorough and the site was untouched.

Why this is not a technical SEO audit

A technical audit reviews how the site is built: crawling, indexing, canonical tags, page speed, structured data. A content audit reviews what the pages say, who they are for, and whether they still deserve to exist. Those are different jobs with different deliverables, and they are usually done by different people.

They touch in one place, which is internal linking, because a page's usefulness often depends on what points at it. Beyond that, keep them separate. One project asked to do both produces a long list of technical tickets and no content decisions.

One piece of sequencing matters. If crawling or indexing is genuinely broken, fix that first, because every measurement the audit depends on will be wrong until the pages can be reached and read.

Start with an inventory, not an opinion

Where the URL list should come from

The inventory starts with a list of every URL that has ever been published, not a list of the pages you remember. Crawl the site, pull the XML sitemap, and export the pages your analytics has recorded. Compare the three, because each one is incomplete on its own.

The differences are the useful part of the exercise. A URL in the sitemap that never appears in analytics is a page with no readers. A URL with traffic that is missing from the sitemap may be orphaned or excluded by accident. A URL in analytics that the crawl cannot reach needs a technical answer before it needs a content decision.

Do not begin by choosing which pages deserve attention. Assemble the full set first, then strip out the duplicates and the non-content URLs such as pagination, tag archives and internal search results.

Fields worth capturing on day one

Keep the first version of the inventory small enough that somebody will actually fill it in. For each URL, capture the page title, the meta description, the main topic, the page type, the last meaningful update, and the links pointing at it.

Two of those fields are basic hygiene and are missing more often than people expect. Per the Web Almanac 2024 SEO chapter, "In 2024, 98% of desktop pages had a title element and 66.7% of desktop pages had a meta description." A third of pages run without a description, and in an audit that is a decision waiting to be made rather than a formatting problem.

Leave performance columns out of version one. If the audit is built around a score nobody trusts, the first argument in the room will be about the score instead of the pages.

Score every page against the job it does

Traffic, links and enquiry signals

Score each page on what it earns, not on how it looks. Organic entrances, backlinks, internal links, and enquiries where you can attribute them. A page that brings in one enquiry a month does more for the business than a prettier page with no readers at all.

This is where expectations need to be realistic. Per Ahrefs research on search traffic, "96.55% of all pages in our index get zero traffic from Google, and 1.94% get between one and ten monthly visits." Most pages on the open web are read by nobody, and most sites have their own version of that distribution. The audit's job is to find the small set of pages that carry the business and the larger set that does not.

Sort by what each page earns and put the top of that list in front of the people who will act on it. Any page with no traffic, no links and no enquiries needs one of two answers: a job it explains for the reader, or a decision to remove it.

Intent fit and commercial relevance

For each page that does earn something, ask whether it is still aimed at the right need. A page written three years ago for a question people no longer ask, or now ask differently, will read as thorough and still underperform.

Commercial relevance is the second half of that. Some pages exist to inform and some exist to sell, and the two need different standards. I want to know which commercial pages the informational ones lead into, and whether that path still works. A page that serves the right intent and sits one link away from the service it supports is doing its job.

The five decisions: keep, improve, merge, redirect, remove

When merging beats rewriting

Merge when two pages genuinely answer the same question. Two thin pages on one topic split their own authority and neither earns a clear position. Combined into a single page that answers the question properly, they usually outperform both originals.

Do not merge pages that mention the same keyword but serve different needs. If one page is for a buyer comparing options and another answers a maintenance question, collapsing them loses coverage and serves neither reader well. Keyword overlap is not intent overlap.

When removal is the honest option

Removal is right when a page has no demand, nothing pointing at it, no enquiries, and no job it could plausibly do. That happens more often than people expect, particularly where several years of campaign pages, event announcements and discontinued services have piled up.

Before you remove anything, check the backlinks and the internal links. A page with no traffic can still hold a link that carries value, or answer a question customers ask on the phone. If it earns nothing and nothing points at it, redirect it to the closest relevant page rather than letting it return a 404.

Judging quality without fooling yourself

Coverage, accuracy and originality

Quality comes down to three questions. Does the page cover the topic well enough that a reader does not need to go elsewhere? Is everything on it still accurate, including dates, prices and figures? And does it contain something a competitor's page does not?

The second question catches more problems than the other two combined. A page from 2022 with a reference to a tool that has since been renamed, a price that has moved, or a year that has passed reads as neglect even when the advice still holds up.

Originality is the one most sites fail quietly. If a page says nothing another page does not already say, it is not adding to the site, and adding more of it will not help. The fix is usually the opposite of writing more: fewer pages, each with something specific to say.

The questions Google's own guidance asks you

Google's public guidance on helpful content sets out self-assessment questions, and they work well as an audit checklist. Would a reader leave feeling they learned enough to act? Does the page show first-hand knowledge rather than a summary of what is already published? Was it written for a reader, or for a figure on a dashboard?

I use those as a conversation starter with whoever wrote the page, because the answer to the last question is usually the most honest one in the room. Pages written to hold a keyword tend to read exactly like that.

Auditing at scale without drowning

Sampling by template instead of by page

Once a site passes a few hundred URLs, page-by-page scoring stops working. The answer is to group pages by template and by purpose, then score the group and sample inside it. All the location pages share a design and a job, so a judgement about one says something useful about the rest.

The exception is the top of the list, which is never sampled. The pages that earn the most traffic, links and enquiries deserve to be read by hand, one at a time. Per Ahrefs guidance on running a content audit, "I recommend following the 80/20 rule and auditing the top 20% of pages with the most organic traffic or backlinks." I work the same way. Audit the fifth of the site that does the earning first, act on it, and only then look at the rest.

That order matters because it produces decisions early. A team that spends two months scoring filler pages has nothing to show for it and usually loses interest before reaching the pages that matter.

Setting a cadence you will keep

A full audit once a year and a lighter review once a quarter is realistic for most businesses. The quarterly pass looks at pages that have declined, pages whose facts have gone stale, and anything published since the last review.

Some of this habit is already common, which is encouraging. Per Semrush's content audit guide, "Our 2023 State of Content Marketing report indicates that one-third of marketers conduct content audits twice a year." A third of marketers being that organised leaves plenty of room to be on the right side of it.

The cadence belongs in whoever owns the site. An audit with no recurring date is a project, and projects end.

Recording decisions so they actually happen

Owners, dates and a redirect map

Every decision needs three things: what happens to the page, who does it, and when it will be finished. Without those three, the row is a note rather than a plan, and the note will still be sitting there next year.

The redirect map is a separate document, and it is the piece teams skip most often. Every removal or merge needs a line recording the old URL and the page it now points to, so the work can be executed in one pass and checked afterwards.

Keep the list short and visible. Twenty decisions that get made beat two thousand rows that do not, and the difference shows on the site within a month.

What to re-measure afterwards

Once the changes are live, measure the pages you touched against a baseline captured before the work started. Traffic to the pages that were improved, and to the pages that received redirected traffic. Enquiries, where you can attribute them. Errors, which should stay at zero.

I would give a refresh six to twelve weeks before judging it, because a rewritten page can dip before it settles. What matters at the end is whether the same decisions produce the same reasoning next time, so the audit becomes a routine rather than a rescue.

The mistakes I see in almost every content audit

Auditing everything and changing nothing

The most expensive mistake is treating completeness as the goal. If the audit's success measure is that every URL has been scored, then the work can be finished without anything changing on the site, which is a strange definition of done for a piece of work about content quality.

I would rather see one afternoon spent on the twenty pages that earn most of the traffic, ending with fifteen decisions and a redirect map, than six months spent producing a document nobody acts on. Speed here is a feature, not a compromise.

Cutting pages that earn links and enquiries

The opposite mistake is removing pages because they look untidy. A short, plain page that brings in enquiries every month is not clutter. It is one of the few assets on the site doing measurable work, and no design opinion outranks it.

Where the wider picture needs reviewing at the same time, a wider marketing audit can cover the channels around the site in the same pass. If the audit depends on data you cannot reach, getting the data your audit needs is the sensible first step, because an audit built on partial numbers produces partial decisions.

Auditing the top fifth of your pages first is the version of this that actually changes a site. If you want that carried out properly rather than as a spreadsheet exercise, that is what my content marketing services cover, and the questions people ask before starting are answered in the common questions about working with me.

Frequently asked questions

How long should a content audit take?

The inventory can be built in an afternoon from a crawl and your sitemap. The scoring and the decisions are what take time, which is exactly why I would audit the pages that earn the most traffic and links first rather than treating every URL as equal. A focused audit that produces twenty decisions beats a spreadsheet of two thousand rows nobody reads.

Do we need paid tools to run a content audit?

No. A crawl of your own site, your sitemap, a Google Analytics or Search Console export, and a spreadsheet will carry a small or mid-sized site. Paid tools make the inventory faster on large sites, but they do not make the decisions, and a tool export is not an audit.

How do we decide between improving and deleting a page?

Look at what the page still earns and what it was built to do. If it has traffic, links or enquiries and answers a real question badly, improve it. If it has no demand, no links, no enquiries and no clear job, deletion or a redirect is the honest option, provided you check for backlinks and internal links before you remove it.

Should we merge pages or fix them individually?

Merge when two pages are genuinely answering the same question and splitting each other's authority. Do not merge pages just because they mention the same keyword, because two pages serving different intents will both lose if you collapse them into one.

How often should a content audit be repeated?

I would run a full audit yearly and a lighter review quarterly, looking at pages that have declined and pages that have gone stale on facts or dates. Search demand moves, competitors publish, and Australian businesses change what they sell, so a one-off audit is a snapshot rather than a habit.

Can a content audit reduce organic traffic?

It can if it is done carelessly. Removing a page with backlinks and no redirect throws away what it earned, and consolidating two pages that serve different intents loses coverage. That is why every removal in a proper audit is recorded with a redirect decision and re-measured afterwards.