Most content audit guides stop at the point where the work actually gets hard. They tell you to build an inventory, pull your traffic numbers, and then decide whether each page is worth keeping. The deciding is the whole job, and that is the part left to instinct.
We read the three guides currently ranking for this query before writing. Between them they run roughly 11,000 words, and not one of them gives you a number that separates a page you should update from a page you should delete. The best of them asks 17 good diagnostic questions and leaves every answer to your judgement.
This page gives the thresholds. They are ours, drawn from auditing our own blog and our customers' sites, and they are stated plainly enough that you can disagree with a specific number instead of with the whole approach.
What a content audit actually decides
An audit produces one of five decisions per page: keep as is, update, merge into another page, redirect, or delete. Everything else in the process exists to support that choice.
This matters because a lot of audit work goes into gathering data nobody acts on. If a metric will not change one of those five outcomes, you do not need to collect it.
The five outcomes are not equally common. On a neglected blog of 200 or more posts, expect roughly half the pages to need nothing, a quarter to be worth updating, and the rest to be merge, redirect or delete candidates. Working through an update pile that size by hand is the usual reason audits stall, which is the job Distribb’s AI content writer is built for.
Why the standard advice stalls
The guides that rank all describe the same shape of process, and they all get vague in the same place. Ahrefs asks whether the page "could get more organic traffic". Orbit Media asks you to gather colleagues and evaluate together. Mailchimp says to analyse the results.
None of that is wrong. It is just not executable at 3pm on a Tuesday with 240 URLs in a spreadsheet and no strong opinion about row 118.
The reason these guides stay vague is defensible: the right threshold genuinely depends on your site's size and authority. That is an argument for stating your assumptions and giving a number anyway, which is what we do below.
How we built these thresholds
We audited our own blog of roughly 430 posts, plus a set of customer sites ranging from 40 to 900 pages. For each page we recorded impressions, average position, age, and referring domains, then recorded the decision a human made and why.
The thresholds below are the cut points where the human decision flipped consistently. They are rules of thumb from that exercise rather than a validated model, and they assume a site with modest authority publishing in a competitive niche.
Two caveats worth stating. Sites with high domain authority should raise every impression threshold, because their weak pages still collect more impressions than a small site's strong ones. And any page earning links or revenue overrides the table entirely.
Step 1: Build the inventory
You need every indexable URL, not just the ones you remember publishing. Old category pages, tag archives, and abandoned landing pages are where the worst content usually hides.
Crawl the site with Screaming Frog or any crawler that exports to CSV. The free tier covers 500 URLs, which is enough for most blogs.
Export your XML sitemap separately and compare the two lists. Pages in the crawl but not the sitemap, and pages in the sitemap that the crawl never reached, are both worth a look before you go further.
Step 2: Attach the four numbers that matter
For each URL, pull four things. Impressions over the last 90 days and average position both come from Search Console. Publication or last-updated date comes from your CMS. Referring domains come from a backlink tool.
Ignore sessions and bounce rate at this stage. They tell you what happened after someone arrived, which is a content quality question you will answer later, and they will not change any of the five decisions.
Ninety days is the right window because it survives seasonality without letting a page's ancient history distort the picture. If your traffic is strongly seasonal, use 12 months and note which quarter each page peaks in.
Step 3: Apply the thresholds
Work down the table in order and stop at the first row that matches. The order matters, because a page with backlinks and no traffic is a redirect candidate, never a delete.
| Condition | Decision | Why |
|---|---|---|
| Any referring domains from real sites | Keep or redirect, never delete | Deleting discards earned authority |
| Position 1 to 10, impressions rising | Keep as is | Working, leave it alone |
| Position 4 to 15, impressions above 200 | Update | Closest thing to free traffic on your site |
| Position 11 to 30, impressions above 500 | Update, and check intent match | High impressions at low position usually means format mismatch |
| Another page targets the same query better | Merge, then redirect | See the cannibalisation step below |
| Under 20 impressions, older than 6 months, no links | Delete or redirect | It has had time and got nothing |
| Under 20 impressions, younger than 6 months | Wait | Too early to judge |
| Zero impressions, older than 12 months, no links | Delete | Nothing to preserve |
| Off topic for the site, any metrics | Delete or move off the domain | Dilutes topical focus |
The six-month line is the one people argue with, and fairly. A page can take longer than that to rank on a low-authority domain. If your site is new, push it to nine months before you delete anything.
Step 4: Check cannibalisation before you touch anything
This step goes before updating, not after, and skipping it is the most expensive mistake in a content audit. If two of your pages target the same query, updating the weaker one is wasted work, and updating both makes the problem worse.
Group your URLs by the query that brings each one the most impressions. Any query with two or more of your pages against it is a cannibalisation candidate.
Then check which page Google actually prefers by searching the query with a site: filter for your domain. The page Google returns first is the survivor. Merge the others into it rather than picking the one you like better.
We found this on our own blog and it was not a small effect. Roughly one in eight of our posts was competing with a sibling, and in several clusters the page we assumed was the strong one was not the page Google was ranking.
Step 5: Decide what to do with the keepers
Most pages come out of the table as keep, and keep is not the same as ignore. Sort the keepers by impressions and look at the top 20 rows.
For each, check whether the page answers the query it ranks for in the first screen. A page sitting at position 6 with good impressions and no clicks is usually failing at the title and opening, not at the body.
Anything below the top 20 keepers genuinely needs nothing. Write the date in a column and move on.
Auditing for citation, not just clicks
This is the part missing from every guide currently ranking, and it changed how we run audits. A meaningful share of search now ends without a click, and pages that rank well can send almost no traffic while still being read by an answer engine.
Our own Search Console data shows queries with 14,000 impressions at an average position of 8.5 and zero clicks. Under the thresholds above, on impressions alone, those pages look healthy. On clicks alone they look dead. Both readings are wrong.
The practical adjustment is to treat conversational queries differently. If a page's top queries are phrased as questions, judge it on impressions and position rather than clicks, and audit its content for whether a machine can lift a clean answer out of it.
That means direct answers near the top of each section, specific numbers rather than ranges, and claims that stand alone without the surrounding paragraph. A page written that way keeps its value even when the click never happens.
What each action is worth, and how long it takes
Audit findings pile up faster than anyone can act on them, so it helps to know which actions pay first.
| Action | Effort per page | Time to see a result | Typical size of result |
|---|---|---|---|
| Fix title and meta on a page at position 4 to 10 | 15 minutes | 1 to 3 weeks | Small but immediate |
| Update a page at position 11 to 20 | 2 to 4 hours | 4 to 10 weeks | Largest single win |
| Merge two competing pages | 2 to 3 hours | 3 to 8 weeks | Large where cannibalisation was real |
| Redirect a dead page with links | 5 minutes | 2 to 6 weeks | Small, compounds site wide |
| Delete a dead page with no links | 2 minutes | None directly | Crawl efficiency only |
| Rewrite a page from scratch | 6 to 10 hours | 2 to 6 months | Highest variance, often not worth it |
Do the title and meta fixes first. They are the cheapest row on the table and they resolve within weeks, which gives you something to show before the slower work lands.
What breaks when you prune wrong
Deleting content is the only irreversible step in an audit, and there are three ways it goes wrong.
The first is deleting a page that had backlinks. The authority those links carried is gone unless you redirect. Always redirect rather than delete when referring domains exist, even to a loosely related page.
The second is orphaning. When you delete a page, every internal link pointing at it becomes a 404. Crawl the site again after any batch of deletions and fix what broke.
The third is redirect chains. If you redirect page A to page B, then later redirect B to C, you have a chain that leaks signal and slows crawling. Point A at C directly when you make the second redirect.
Doing this without paid tools
You can run a complete audit on free tools, and for sites under about 300 pages the result is close enough to identical.
Search Console gives you impressions, position, and queries per page at no cost. Screaming Frog's free tier crawls 500 URLs. A spreadsheet handles the inventory and the decision column.
The one thing you genuinely cannot get free is referring domains per URL at any scale. Free backlink checkers cap results low enough that they will miss links on exactly the low-traffic pages where the answer changes your decision. If you have no backlink data, redirect rather than delete whenever you are unsure.
How often to re-run it
A full audit once a year is enough for most sites. Anything more frequent and you are measuring noise, because the update actions from the last audit have not had time to resolve.
Two lighter checks are worth running in between. Quarterly, pull every page at position 4 to 15 and fix titles. Monthly, check that newly published posts have internal links pointing at them, because new posts land orphaned by default.
The monthly check matters more than it sounds. On a blog publishing 80 posts a month, a year of neglect leaves several hundred pages with no internal links, and no audit threshold will save a page nothing links to.
A worked example
Here is what the process produced on a 240-post blog we audited this year.
The crawl found 261 indexable URLs, 21 more than the CMS reported, mostly old tag archives. Search Console had data for 198 of them, so 63 pages had never received a single impression.
Applying the table: 104 pages came out as keep, 38 as update, 22 as merge into 9 surviving pages, 41 as redirect, and 56 as delete. The 63 zero-impression pages split between delete and wait depending on age.
The 38 updates were sorted by impressions and worked in that order. The first ten, all sitting between positions 11 and 18, accounted for most of the traffic gained over the following three months. The remaining 28 have still not been touched, which is normal and fine.
Where Distribb fits
We built Distribb around the part of this process that repeats: finding the pages worth updating, checking which ones compete with each other, and rewriting them. It pulls your Search Console data, groups pages by the query they actually rank for, and flags the cannibalisation clusters that the manual site: search finds one query at a time.
The honest limitation is that Distribb does not crawl your site. It works from Search Console and from content it can reach through your CMS, so the inventory step above still needs a crawler, and pages that have never received an impression are invisible to it. For the first pass on a neglected site, run Screaming Frog first and bring the list with you.
If you want the smaller version of this, our AI SEO audit runs the position and cannibalisation checks against a connected Search Console property without the full inventory step.
Related reading
Once the audit tells you which pages to rewrite, the question becomes how to write the replacement. Our guide on how to write content for a website covers the structure work, and if the pages you are auditing were produced by an AI tool, auditing automated content for SEO covers the specific failure modes that come with it.
Frequently asked questions
How long does a content audit take? For 200 to 300 pages, budget one full day for the inventory and decisions, then treat the actions as a separate project. The audit itself is fast. Acting on it is what takes months.
Should I delete or noindex low-quality pages? Delete and redirect if the page has links, delete outright if it does not. Noindex leaves the page live and still consuming crawl budget, which solves half the problem and adds a thing to maintain.
Does deleting content hurt rankings? Deleting pages that have backlinks or traffic does. Deleting pages with neither has no negative effect and usually a small positive one, because it concentrates your topical signal.
What if a page ranks well but gets no clicks? Check whether the query is conversational. If it is, the page is probably being read in an AI answer and the click was never available. Judge it on impressions and position, and make sure the answers in it are liftable.
Can I audit a site I do not have Search Console access to? Not properly. You can crawl it and assess quality by hand, but without impression and position data you are guessing at which pages matter. Get access first.
What to do next
Crawl the site, pull 90 days of Search Console data, and put both in one spreadsheet with a decision column. Work down the thresholds table until every row has a decision, then sort by impressions and start with the title fixes.
The audit is worth nothing until the actions ship. If you want the update and cannibalisation half handled automatically against your own Search Console data, start with Distribb.