Free template
The page nobody visits might be the one ChatGPT quotes.
Every content audit ever written deletes it. Zero clicks, no links, nobody cares. But engines quote pages, not rankings, and the moment you delete that page you leave the answer permanently. This one checks before it cuts.
No email. No signup. Rebrand it and use it with your own clients if you like.
How the file decides. The order is the argument.
1
Is an AI engine citing this page?
Keep
2
Is it converting? Leads, calls, forms.
Keep
3
Is another page of yours fighting it for the same query?
Pick one
Every other content audit starts here
4
Is it earning clicks?
Keep
5
Impressions, but ranking badly?
Update
6
None of the above?
Retire
Traffic is the fourth question, not the first. A cited page with zero clicks passes at step one and never reaches the part of the file that would have killed it.
The problem
Pruning advice was written for a web that no longer exists.
The rule everyone still uses: no traffic, no links, no rankings, delete it. It was good advice. It was written when the only way a page could help you was by being clicked, and that stopped being true about two years ago.
Engines quote pages, not rankings
A clean explainer sitting at position 40 that nobody has ever visited can be the source an assistant reaches for when someone asks about your category. It earns you nothing in analytics and it is the reason you get named in the answer. Delete it and you leave the answer, permanently.
Those pages look exactly like dead ones
That is the trap. Zero clicks, low impressions, no backlinks, old. Every test ever written for pruning marks it for deletion, and the page fails all of them while quietly doing the most valuable job on your site. There is no way to tell the difference except to look.
So AI citation is checked first, and it wins
Not weighted alongside traffic. Checked before it. If an engine is citing the page it stays, whatever the analytics say, and the file will not let you argue. That single rule is the whole difference between this and the audit template you may already have.
The file
One row per URL. Four bands.
PAGE, PERFORMANCE, DISCOVERY, DECISION. Export the URL list from your sitemap rather than from memory, because the pages you forget you published are exactly the ones this is for. Most of the columns are the usual ones. These are not.
KAI Cited
Has an engine actually quoted this page. Yes, No, Unknown.
The column that makes this file different, and the only one that overrides everything else. Check it in ChatGPT, Perplexity, Gemini and AI Overviews before you touch a single Verdict. There is no tool that does this properly, which is precisely why almost nobody does it, which is why the pages get deleted.
YesNoUnknown
CCluster
The query this page actually targets. Not the title.
The most important thing you type, and the grouping key for everything else. Two pages sharing a Cluster are competing with each other. Be honest: if two pages differ by the word “in”, they are the same cluster. In Cluster counts them for you and turns brand red above one.
ELast Real Update
The last time a human meaningfully changed the page.
Not the CMS modified date. A bulk re-save touches every page and changes nothing, and if you record that date you will believe your content is fresher than it is. If you cannot remember editing it, it has not been edited.
JConversions
Leads, calls, form fills. Whatever counts for this business.
The column most content audits skip entirely, and the one that saves the ugly old page nobody is proud of that quietly books three calls a month.
OSignal
The file’s suggestion, calculated from everything to its left.
A starting point for the argument, not the end of it. It reads in the order at the top of this page: cited, converting, cannibalising, earning, ranking, dead. Override it whenever you have a reason and put the reason in Notes. The sheet does not know your business.
Cited: keepCannibalising: pick oneRanking badly: updateDead: retire
P / QVerdict and Destination
Keep, Update, Consolidate, Redirect, Noindex, Retire.
Your decision, and where it goes. Required for Consolidate and Redirect: a redirect without a decided destination becomes a redirect to the homepage, which is deleting the page with extra steps.
KeepUpdateConsolidateRedirectNoindexRetire
The scorecard
Three numbers that stop you.
The scorecard counts the usual things: pages, verdicts, clusters, index reduction. These three are different. They exist to catch you in the act, and they are the most useful part of the file.
Cited pages you are about to delete
Should always be zero. Anything here is the file contradicting its own rule, and it means you overrode a Cited: keep without noticing.
Conversions you are about to delete
Should be zero. If it is not, stop and re-read those rows. You are about to remove something that was making money because it was ugly.
Clicks you are about to delete
If this is not close to zero, you are not pruning, you are cutting. Redirect them instead. A 301 costs nothing and keeps whatever the page had.
And one that tells you whether it was worth doing
Cited with zero clicks. Every one of those pages would have been deleted by a traditional audit, and every one is a source an engine reaches for. It is the headline number on the scorecard, and on most sites older than two years it is not zero.
The method
How to run it yourself.
The whole thing is on this page. If you want to do it yourself, this is everything you need.
1
Pull every indexable URL from the sitemap
Not from memory, not from the nav. The pages you have forgotten about are the entire reason for doing this, and they are the ones you will not think to list.
Then add the numbers: clicks, impressions, position, referring domains, conversions. Search Console for the first three, and use Search Console rather than a rank tracker, because the tracker is modelling and Search Console is counting.
2
Write the Cluster for every row before you look at anything else
One query per page, the one it actually targets. This is the slow part and it is where the value is. When you finish, In Cluster tells you how many pages are fighting each other, and on an old site the answer is usually four to six on your best query.
Do this before the numbers seduce you. Once you know which page has the traffic you will unconsciously write clusters that justify keeping it.
3
Check AI Cited. Manually. Before any Verdict.
Run the page’s topic as a real question in ChatGPT, Perplexity, Gemini and Google AI Overviews, in a clean session with no chat history. If your page turns up as a source, mark it Yes and write the prompt in column M.
It is about a minute per page and there is no tool that does it properly. Do it before you set a single Verdict. Once you have decided a page is dead it becomes very hard to notice that an engine disagrees.
4
Read the Signal, then overrule it
It is a formula. It has never met your customers. Where you disagree, disagree, and write why in Notes so that the next person through the file, probably you in eight months, can see the reasoning rather than guess at it.
Where a cluster is flagged, pick the winner deliberately: usually the page with the traffic, but if a different one is the one being cited, that is your winner instead.
5
Redirect, do not delete. Then read the three safety numbers.
A 301 keeps whatever equity the page had and costs nothing. Retire only when there is genuinely nowhere sensible to point. Consolidate means the content moves, not just the URL: merging six thin posts into one thin post is not consolidation, it is a longer thin post.
Then check the three numbers above before you touch the CMS. If cited, converting, or clicks-about-to-be-deleted read anything other than zero, you are not finished arguing with yourself.
Calibration
Three ways people get this wrong.
Deleting instead of redirecting
Deletion is the only irreversible option on the list, and it is almost never the best one. A 301 into the page that survived keeps the links, the history and whatever the old page was quietly doing that you did not measure.
Checking citation after deciding
The order in the file is not decoration. Once a page is in your head as dead, you will read the evidence to fit, and you will find a reason why that one citation does not count. Check first, decide second.
Pruning because a blog post said to
Index bloat is real and it is also the most over-prescribed diagnosis in SEO. If your pages are cited, converting, or holding a cluster on their own, the number of them is not your problem. Cutting a hundred pages to fix a problem you did not have is not a strategy, it is a weekend you will not get back.
Questions
FAQ
Is the content pruning template really free?
Yes, and there is no email form. Download it, edit it, rebrand it, use it with your own clients. It is the file we run content audits on, published as it is.
Why does AI citation override traffic?
Because a citation is worth something a click is not. When an engine answers a question and names your business, that is reach you cannot buy and cannot get back once the source page is gone. Traffic is a good signal and it is no longer the only one, so it is the fourth question in this file rather than the first.
How do I check whether a page is cited?
By hand. Run the page’s topic as a real question in ChatGPT, Perplexity, Gemini and Google AI Overviews, in a clean session with no chat history, and see whether your URL turns up as a source. About a minute per page. There are paid tools that sample this at scale, and for a one-off audit of a few dozen pages, doing it yourself is faster than setting one up.
How many pages should I expect to remove?
On a site nobody has edited in three years, somewhere between a fifth and a half usually ends up redirected, consolidated or noindexed. But if the number comes out low, that is a finding and not a failure. Some sites do not need pruning.
Can I use it in Google Sheets?
Yes. Use the Make a copy link at the top of this page. The dropdowns, colour coding and the live scorecard all carry across.
How many pages does it handle?
The formulas and dropdowns run to row 205, so 200 URLs. Add rows past that and you have to extend both, or the scorecard will quietly stop counting them.
Or we can just run it
The spreadsheet is the easy part.
The hard part is the argument about which page survives, and having it with someone who has no attachment to any of them. We have done this before and written it up: the content pruning case study is the honest version, including what it cost.
