Content health audit: which post should you update?

It is 9:40 on a Tuesday morning. Before opening the quarterly plan, a content editor pulls up Search Console, lines up the last six months against the six before them, and finds that 31 of the blog's 214 posts have lost more than half their clicks. She has one writer, two weeks and 31 candidates. The list is long and the capacity is short. The real question is not "should we refresh content" — it is "which six of these 31 rows deserve the next two weeks?"

This article describes how to make that choice with a repeatable scan instead of a gut feeling. We start with why not every declining page deserves an update, then look at the signals a content health scan collects. From there we separate a genuine decline from seasonality and lost demand, find the pages competing with each other for the same query, rank the candidates with a table, and settle the update-merge-redirect-remove decision. At the end we cover what actually changes during an update round, how to measure whether it worked, and who owns the scan on what cadence.

Why doesn't every declining post deserve an update?

A traffic drop is not automatically a defect. A post can decline because nobody searches the topic any more, and it can decline because the intent behind the query has shifted. If the "how to set up this integration" guide you wrote in 2022 goes unread because the integration interface was rebuilt from scratch, the problem is not the prose — it is the topic. Refreshing that page spends two writer days on work that will never come back.

Updating has a real cost: research, drafting, editing, visuals, publishing and re-indexing. That cost only pays back when three conditions hold together. There is still demand for the topic; the page as it stands does not satisfy that demand; and the page does a job for the business, meaning it connects to a product, a demo or a sign-up. When all three are missing, the right decision is not to update but to drop the row. When a refresh round is worth running and how to structure it is covered at length in our guide to refreshing older content; the question here sits one step earlier — which page gets into that round at all.

What signals does a content health scan actually read?

A health scan is not a traffic report. Traffic is one signal and on its own it misleads. The minimum set to collect looks like this: impression and click trend, average position band, the queries a page appears for, how many of your pages surface for the same query, status code and indexability, the number of internal links pointing at the page, click depth from the homepage, the last modified date, and the page's contribution to conversions.

Inside Rocketly's SEO Center, the Content Health component takes on the two hardest parts of that set: it lists declining pages and keyword cannibalisation — pages competing with each other for the same query — before the lost clicks show up in a report. The technical half sits in Page Detail, where status code, indexability, response time, incoming internal links and click depth appear for each URL on a single row. So while you are asking "why did this post drop", you notice that it was never indexed, or never received a single internal link, without switching tools. Signals should be read in clusters rather than one at a time — the same logic we set out in the piece on topic clusters and pillar pages applies here: a post is judged inside the cluster it belongs to, not in isolation.

Decline, seasonality or lost demand?

The same downward chart can have three different causes, and each calls for a different answer. The fastest way to separate them is to read clicks alongside impressions. If impressions hold steady while clicks fall, the page is still being shown and simply is not being clicked; the culprit is usually the title and description, or a new element that has appeared above the results. If impressions and clicks fall together, either you lost position or demand for the query shrank.

For the third distinction, compare year over year rather than month over month. November being below October carries no information in most industries; November being below last November does. It is also worth accounting for changes in the result page itself: the AI Summary Check in the SEO Center looks at whether an AI overview appears for a query and lowers the expected click gain accordingly. That small check explains how you end up with "our position is unchanged but clicks halved". On the production side, how AI-assisted drafting affects the text and how search engines treat it is discussed separately in our assessment of AI-generated content.

How many of your pages compete for the same query?

This is the problem a health scan finds most often and teams overlook most easily. Over three years you publish four separate posts around the same topic; none of them is bad, but the search engine cannot decide which one to show. The result is four pages drifting between position seven and fifteen. One strong page could reach position six; four weak ones reach nowhere.

The symptoms are familiar: impressions for a query are split across several URLs, the page that surfaces changes from week to week, and none of them settles into a stable slot. To decide, first identify which page has the highest clicks and the best average position for that query — that page is the canonical candidate. The others either need to be narrowed so they serve a genuinely different intent, or merged into the canonical page and redirected. How to draw that line, and the difference between merging and narrowing, is worked through with examples in the article on keyword cannibalisation.

Which post goes first? A prioritisation table

Once the candidates exist, the ordering should not be up for debate. The table below turns the signal combinations from the scan output into decisions. The numbers are example thresholds; you shift them to match your own site's volume, but the logic stays the same.

Observed state	Signal combination	Decision	Priority
Stuck on page two	Average position 11–20, high impressions, low clicks	Expand and add internal links	Highest
Quiet erosion	Position held, clicks down year over year	Rewrite title and description	High
Split query	3+ URLs for one query, none stable	Merge and redirect to one page	High
Technical block	Not indexed, orphaned or returning 4xx	Fix the technical issue first	Medium but fast
Demand gone	Impressions and query count both near zero	Archive or remove	Low

Why is the first row the highest priority? Because the gain sits closest. The Quick Wins list in the SEO Center produces exactly that row: queries stuck on the second page are ranked by the estimated clicks you would gain by moving them onto the first. Ordering should also be matched against business impact — reading lead source analysis and attribution helps here, because the page that earns the most clicks and the page that earns the most sign-ups are rarely the same page.

Update, merge, redirect or remove: which decision fits which case?

There are four decisions and every candidate has to land on one of them. An update is right when the page's intent is still correct but its content has aged: the screenshots have changed, the steps are different, the examples describe an older period. A merge fits when two or more pages partially serve the same intent; the strongest page becomes the base, the distinctive sections of the others move into it, and the old addresses point to the new one with a permanent redirect.

A redirect on its own is for cases where there is nothing worth carrying over: the page is weak but has earned links. Removal is the least used and most needed decision of the four. Campaign announcements, past event pages and guides to products you no longer sell keep polluting the scan list and get re-evaluated every round. When deciding, ask what role the page plays in the business: does it answer a demand, or is it simply sitting in the archive? We set out how content earns its place in the pipeline in the piece on content marketing for lead generation, and the same test applies during a health scan.

What actually changes during an update round?

In many teams "we updated it" means the date was changed, which changes nothing. A meaningful round does at least four things: the subtopics that today's result page covers are added to the text, outdated sections are deleted, examples and screenshots are refreshed, and the title and meta description are rewritten for the current intent. Add to that a review of the internal links pointing at the page; if the updated post receives no links from the rest of its cluster, the update is half done.

The writer should run this round against a defined task rather than intuition: which queries are targeted, which sections are added, which are cut, which pages get linked. Committing that list to a template stops every update round from becoming its own negotiation, and a content brief template exists to describe exactly that task.

Internal linking does not have to be done blind either. Internal Link Opportunities in the SEO Center ranks pages that receive few internal links or sit deep in the link graph, and leaves the change to human approval. It also matters increasingly that an updated page can be cited — that it is clear who wrote it, what it is based on and when it was verified. We went through those signals in the article on which brand AI systems cite.

How do you know whether the change worked?

The step teams skip most often is measuring the result. Once the update is live, two things are needed: the date of the change has to be recorded, and the before and after have to be compared across two windows of equal length. In practice four-week windows work well; one week is too noisy, three months lets other changes creep in. The metrics to compare are clicks, impressions and average position, with conversions as a fourth column where you have them.

In the SEO Center this runs through Change Impact Measurement: you mark the fix you made, and the field data from the periods before and after is placed side by side. "We changed the titles — did it work?" is then answered by comparison rather than by feel. If you want to see a page's crawl, index, redirect and removal history on a single timeline, URL Lifecycle answers the same question along the time axis. Rather than waiting for updated pages to be re-crawled, you can notify engines that support IndexNow through Instant Index Notification; Google does not use the protocol, so waiting times there follow their usual course.

Recording the outcome of a measurement round matters as much as the measurement itself. Which kind of update works on your particular site only becomes visible after five or ten rounds have accumulated: on some sites title rewrites dominate, on others adding sections does. That accumulation lets the next quarter's priority table rest on your own history rather than on guesswork.

On what cadence does the scan run, and who owns it?

A health scan is a rhythm, not a project. Quarterly is enough for a small blog; monthly suits a team publishing several posts a week. What usually breaks the rhythm is a lack of ownership: the scan runs, the list appears, and because the list never lands in anyone's task queue it slips to the next quarter. That is why the output has to feed a queue. The SEO Operations work queue in the SEO Center collects findings into one place; each row moves with an owner, a priority and a status, and closes when the fix is verified in the following crawl. The Health Score summarises the difference between rounds: every scan lists new, resolved and ongoing issues separately.

Update work should not sit apart from the production plan either; new posts and refreshes need to compete in the same calendar so capacity splits realistically. How to set up that split is covered in the piece on the content calendar and publishing plan. Another common finding during a scan is the strong but tired page: the writing is good, the topic is alive, but the format has aged. For those, a full rewrite is often more expensive than it needs to be, and repurposing the existing content delivers a faster result.

If you want the audit, rank tracking and fix queue in one panel, open a free Rocketly account and run the first scan against your own site.

FAQ

How often should I run a content health scan?

It depends on your publishing pace. If you publish a few posts a month, one round per quarter is enough. At several posts a week, a monthly rhythm fits better, otherwise the candidate list grows beyond what a single round can process. Technical findings can be crawled more frequently; for content decisions, human capacity is the limiting factor.

How many posts should I update at once?

As many as your capacity allows, without breaking the measurement. If you change thirty pages in the same week, you cannot tell which change produced which result. Working in small batches with a measurement window between them is both more manageable and easier to learn from.

Will deleting a declining page hurt the rest of my site?

Removing a page with no remaining demand, no inbound links and no contribution to conversions is generally safe. The risky move is quietly removing a page that has earned external links or still produces impressions. In those cases, redirect it permanently to the strongest page on the nearest topic instead of deleting it.

How long does it take to see the effect of an update?

It varies from a few days to a few weeks, because the page has to be re-crawled and re-evaluated. That is why taking a four-week measurement window is more reliable than reacting to the fluctuation of the first few days.