Content Audit: Keep, Update, Merge or Remove

Four verdicts and no fifth. An audit without a decision column is a spreadsheet, not an audit.

An audit is not an export. You can pull forty columns per URL and still have nothing to do on Monday.

So build it backwards. Decide what the four possible outcomes are, then collect only the data that separates them. This is the cleanup half of our content marketing guide; the planning half lives in the mapping guide.

Everything below assumes you own the site and have Search Console and GA4 access.

What a content audit is, and the four verdicts

A content audit is the inventory of what actually exists on your site, judged against what each page was supposed to do. It is the opposite end of the same job as content mapping — the map is the target state, the audit is the reality, and the gap between them is the finding.

Four verdicts. No fifth, and no "revisit later" column, because that column is where audits go to die.

Note the second row. An update happens on the existing URL. Republishing a refresh at a new URL throws away whatever the old one had earned — the most common self-inflicted wound in this whole exercise.

The inventory: the columns you need, and where each comes from

One row per URL, eleven columns, not forty. Each exists because a verdict depends on it; anything that does not change a verdict is decoration.

One note on Search Console. Its queries dimension "groups your data by the search query users typed", and the default Performance report view shows only "the past three months" — so widen the date range before you judge anything, per Google's Search Console Help on the Performance report (accessed 25 September 2026; Google publishes no last-updated date on its Search Console Help pages, so none can be quoted).

The last column is the one that matters. A page with no stated job cannot be defended. A page with a clear job and no traffic is usually an update, not a deletion.

The decision rules, in order

Run these in sequence and stop at the first one that matches. Order is what removes the judgement calls — the same URL handed to two people should come out with the same verdict.

1. Published less than six months ago? Keep. Organic performance typically takes about three to six months to mean anything, so the page has not had its chance.
2. Not indexed? Keep, and raise a technical ticket instead. Zero clicks on an unindexed page says nothing about the writing.
3. Has referring domains, or conversions, or both? Keep or update — never remove, never merge away. Traffic is one reason a page exists; being linked to is another.
4. Does another of your URLs rank for the same top query with the same intent? Merge. That test is the next section, and it is the most valuable thing an audit finds.
5. Ranking roughly positions 5–20 for a query you want, or clearly out of date? Update. The demand and the signal already exist; the page is not yet the best answer.
6. None of the above, and you would not brief this page today? Remove.
7. None of the above, but you would still brief it today? Update — and fix the title and internal links first, as the cheapest possible intervention.

The honest problem with traffic thresholds

Every audit template sets a cut-off: under ten clicks in six months, delete. That number is arbitrary, and so is every replacement for it.

It has to be. Ten clicks is a failure for a page chasing a 5,000-a-month query and a success for one chasing a 30-a-month query that closes deals. Clicks depend on the query's size, the page's age and season, its internal links, and whether it was ever indexed. One number cannot carry that.

So stop hunting for a better threshold. Ask three questions instead:

• What was this page for, and is anything else doing that job better? If yes, that is a merge, not a deletion.
• What breaks if it disappears? Inbound links, internal links, a live campaign, a sales team that sends it to prospects.
• Would you commission it today? The replacement test is the only one that scales, and it is at least honest about being a judgement.

Use a threshold to sort the queue, never to decide. Low traffic buys a page a review, not a sentence.

Cannibalisation: two of your pages, one query

The single most valuable finding in most audits is two URLs competing for the same query, because fixing it costs nothing and needs no new content. Here is how to find it in Search Console without a tool.

1. Set the Performance report range to 12 months. The default three-month window hides the flipping you are looking for.
2. Filter by one query, then switch to the Pages tab. Two or more of your URLs showing impressions for that single query is a candidate.
3. Compare two six-month windows for that query. If the URL earning the impressions changes between them, Google is unsure which page answers it. That instability is the symptom — not the mere presence of two URLs.
4. Read both pages and name the question each answers. Same question, same intent, two URLs — merge. Different questions sharing a phrase — differentiate instead.
5. Check where you link internally. Often both pages survive because half your internal links point at one and half at the other. Pick one and re-point the rest.

Two of your URLs appearing for one query is not automatically a fault. It becomes one when neither ranks well and the winner keeps swapping. Whether two pages share an intent is a call you make with our search intent guide; grouping the survivors is what topic clusters are for.

Three ways a content audit goes wrong

All three are recoverable except the second, which is why it gets the most space.

Google's John Mueller has pushed back on mass deletion for years. On X, on 22 September 2023, he wrote: "keep in mind that removing content doesn't make the rest rank higher" (reported by Search Engine Roundtable, 24 September 2023). He has not ruled pruning out either — on CNET's plan to remove old pages he said he thought "their plans were pretty reasonable & nuanced", while adding that "the SEO effects are unlikely to be noticeable" (reported by Search Engine Roundtable, 14 August 2023).

Read together, that is the honest position: removal is housekeeping, not a ranking tactic. Delete pages because they are wrong, thin or embarrassing — not because you hope the survivors move up.

Executing each verdict

Keep needs nothing but a date. The other three need doing properly, and two of them are technical.

• Update means rewriting on the same URL. Write a fresh content brief as though the page did not exist, change the visible last-updated date, and only then ask whether part of it is better off in another format via repurposing.
• Merge means one page survives and the rest point at it.
• Remove means one of four technically different things, and picking the wrong one is how audits break pages people were relying on.

How a merge is actually executed

1. Pick the winner on earned signals, not writing quality. The URL with the better query footprint and the links wins, even if the other reads better. You can rewrite prose; you cannot rewrite history.
2. Move across only the unique substance — the sections the winner does not already have. A merge that concatenates two pages produces something worse than either.
3. Redirect each loser to the winner with a permanent redirect, and to the closest equivalent page rather than the homepage. Google's documentation is explicit that permanent redirects "show the new redirect target in search results" and that "the indexing pipeline uses the redirect as a signal that the redirect target should be canonical", while temporary redirects "show the source page in search results" (Redirects and Google Search, last updated 14 April 2026). A 302 here is the wrong tool.
4. Repoint every internal link at the winner directly, rather than leaving it to travel through a redirect. Mechanics in our internal linking guide.
5. Watch the winner's query set for six to eight weeks. You are checking that the loser's queries arrive, not that traffic instantly rises.

What "remove" actually means: 404, 410, noindex or redirect

Four different outcomes, not interchangeable.

Both status-code rows come from Google's HTTP status codes and network errors documentation, last updated 4 February 2026 — worth reading because it kills the widespread claim that a 410 deindexes faster than a 404. It does not.

noindex has one trap: if the URL is also blocked in robots.txt, "the crawler will never see the noindex rule, and the page can still appear in search results" (Block search indexing with noindex, last updated 10 December 2025). Pick one mechanism, not both. On timing, Mueller said on Reddit that for noindex "there's no specific time, but it's usually less than a few months. I think I've seen it take up to 6 months" (reported by Search Engine Journal, 19 August 2024).

The Removals tool in Search Console is not a removal method. Google's Help page says a temporary removal lasts "only about six months", that "blocking a URL does not prevent Google from crawling your page, only from showing it in Search results", and that it is "only one step in this process to remove a URL permanently" (accessed 25 September 2026; as with every Search Console Help page, no last-updated date is published on it). Use it in an emergency, then do the real fix.

What we keep finding in our own audits

Our own observation, from running audits on client libraries rather than from any study: the remove pile is always smaller than the client expected and the update pile always larger. People arrive wanting to delete a third of the site and leave deleting a handful of pages.

Most "dead" pages are not bad pages — they are pages nobody links to internally, pages competing with a sibling, or pages nobody ever finished. Merges and updates wearing a deletion costume. We have no measured figure for this and will not invent one; it is a pattern we keep meeting, stated as a pattern.

The question we ask in every audit workshop, before any spreadsheet opens: name the page you would send a prospect today. Whatever gets named is what the audit should protect, and it is rarely the page with the most traffic.

If you would rather the audit was run with you and the update pile actually written, that is what our content marketing services do. Directionally, content marketing in India runs about ₹25,000 to ₹1,50,000 a month depending on volume and depth — a market range, not a quote.

Where to go from here

Audit once a year, not once a quarter — a quarterly audit judges pages on a window shorter than the time they need to work.

• Week one. Crawl the site, pull twelve months of Search Console and GA4, and fill the inventory — including the "what was this page for" column, by hand.
• Week two. Run the rules in order. Every URL leaves with one of four verdicts; nothing sits in a maybe column.
• Week three. Execute the merges first — fastest wins, no new writing.
• Then. Turn the update pile into briefs, sequence them against your content marketing strategy, and leave removals last.

One expectation to set with whoever signs this off: consolidating and rewriting typically takes about three to six months to show meaningful organic movement, and some merges dip for a few weeks before the queries settle on the winner. Neither is a reason to reverse it.