On-page mechanics
How to run a content audit on a page you already published
A page-level SEO content audit that ends in one of five decisions — refresh, rewrite, merge, prune or leave alone — instead of a spreadsheet nobody actions.
7 min read
Most content audits die in the spreadsheet. Someone exports every URL, scores them on eight columns, colour-codes the result, and then nothing happens, because "score: 62" is not an instruction.
This process is built the other way round. Every page you audit exits with exactly one of five decisions, and each decision has a defined next action. The data you collect is only the data needed to choose between them.
It works at any scale, but it is designed for sites where a human still reads the pages — say up to a few hundred URLs. Beyond that the diagnostic questions stay the same; only the triage layer changes.
The five outcomes, defined before you start
Deciding what the possible answers are before you look at data is what stops an audit from turning into an opinion exercise. These five cover every published page.
| Decision | When it applies | What it costs |
|---|---|---|
| Leave alone | Performance is stable and the content is still accurate | Nothing — recheck next cycle |
| Refresh | The angle is right; facts, examples or screenshots have aged | An hour or two per page |
| Rewrite | The page targets the right query but answers it badly or thinly | Comparable to writing a new page |
| Merge | Two or more pages compete for the same intent | A rewrite plus a 301 redirect |
| Prune | No traffic, no links, no strategic role, and nothing worth merging | A redirect or a deletion, plus removal from internal links |
Step 1 — Pull the four numbers that matter
You need four inputs per URL. Anything else is optional and mostly slows the audit down.
- Clicks and impressions, last 3 months vs the same 3 months a year ago — From Search Console. Year-over-year comparison rather than month-over-month, because it removes seasonality. The shape of the change matters more than the absolute number.
- Top queries and average position for this URL — Also Search Console, per page. This tells you what the page is actually being matched to, which is regularly not what it was written for.
- Last meaningful update date — Not the CMS timestamp from a typo fix — the date the substance last changed. If you do not track this, the audit is the moment to start.
- Whether another page on your site targets the same query — Search "site:yourdomain.com <query>" or filter your Search Console query export. Two URLs alternating on the same query is the cannibalisation signal.
Step 2 — Diagnose the pattern
Four numbers produce a small number of recognisable patterns. Match the page to one before deciding anything.
| Pattern | What the data looks like | Usual cause | Decision |
|---|---|---|---|
| Decay | Impressions flat or up, clicks down, position slipping slowly | Competitors published something more current; your facts aged | Refresh |
| Intent mismatch | Impressions present, position stuck around 8-20, queries not the ones you targeted | The page answers a different question than the one it ranks for | Rewrite |
| Cannibalisation | Two URLs alternating on the same query, neither in the top 5 | Overlapping pages splitting relevance signals | Merge |
| Thin or orphaned | Near-zero impressions after 6+ months, no inbound internal links | The page was never discoverable or never had a reason to exist | Prune (or link it properly first, then re-measure) |
| Healthy | Stable clicks, stable position, content still accurate | Nothing | Leave alone |
Decay vs intent mismatch — the distinction people get wrong
Both look like "the page used to do better". The tell is the query list. In decay, the page still ranks for the queries it was written for, just lower. In an intent mismatch, the query list has drifted: the page ranks for adjacent questions it only partly answers, which caps its position no matter how much you polish the prose.
Refreshing an intent-mismatched page is the single most common wasted audit action. You update the statistics, republish, and nothing moves, because the page is still answering the wrong question well.
Check whether the SERP itself changed
Before diagnosing anything as your page's fault, look at the live results page. If the format shifted — the top results are now tools, videos or comparison tables where they used to be guides — the problem is format, not quality, and the fix is a different kind of page.
Step 3 — Audit the page itself
Only now open the page. The data told you what is happening; the page tells you what to change. Work through the mechanics first because they are cheap, then the substance.
- Structure: one H1, no skipped heading levels, each section answering its own heading in the first sentence.
- Metadata: title still describing the current content, description still offering a reason to click, canonical self-referencing.
- Coverage: list the sub-questions the current top results answer and this page does not. That list is the refresh brief.
- Accuracy: dated statistics, screenshots of interfaces that changed, broken outbound links, product claims that are no longer true.
- Evidence: claims that were assertions when you published and could now be supported with a source, a number or an example.
- Internal links: at least two contextual inbound links from related pages, with anchor text that names the topic.
Step 4 — Execute the decision
Refresh
Update facts, add the missing sub-sections, replace stale examples, and tighten the intro so the answer arrives sooner. Keep the URL, keep the H1 unless the subject genuinely changed, and update the visible last-reviewed date only if the substance actually changed. Republishing an unchanged page with a new date is a signal you cannot spend twice.
Rewrite
Re-derive the brief from the queries the page currently ranks for plus the intent you want, then rebuild the outline before writing. A rewrite that starts from the existing draft usually inherits the structural problem that caused the mismatch.
Merge
Pick the URL with the stronger history — links and impressions, not the nicer slug — fold the unique material from the other page into it, and 301 the loser to the winner. Then remove or repoint the internal links that used to feed the redirected URL, otherwise every internal path to that content runs through a redirect.
Prune
If a genuinely relevant destination exists, redirect. If not, let the URL return 410 or 404 and remove it from your sitemap and internal links. Redirecting unrelated pages to the homepage is a well-known soft-404 pattern and helps nobody.
Where a tool helps, and where it does not
The diagnosis stage is judgement work: comparing query lists, reading the current SERP, deciding whether an angle is still right. No tool does that for you, and tools that claim to are usually just thresholding a traffic number.
The execution stage is mechanical enough to automate in part. In wordiqo, the free optimizer takes the text of an existing page plus up to ten target keywords — the first is the primary one — and reports the structural and keyword findings before anything is rewritten: whether an H1 exists, whether there are at least two H2s, whether the primary term appears in the introduction and in a heading, per-keyword counts and density, average sentence length and word count.
From there, "Rewrite & enrich" produces a restructured version with a single H1, a heading tree, an answer-first shape and a proposed meta title and description, while "SEO formatting only" leaves your wording untouched and applies structure and emphasis alone. That second mode is the one that matters for audits of pages whose facts are already correct: you keep your sentences and fix the mechanics.
For a page that came out of the audit needing new sections rather than new sentences, the article editor adds them one topic at a time, with a preview of where each new section will be inserted and checks that reject duplicate headings — which is the failure mode of every "add more content" refresh.
Audit cadence
A full-site audit once a year is worse than a rolling audit of a fixed number of pages each month. The rolling version keeps the decisions small enough to act on the same week you make them.
| Page type | Cadence | What usually triggers work |
|---|---|---|
| Money and commercial pages | Quarterly | Product changes, competitor repositioning |
| Evergreen guides | Twice a year | Decay, missing sub-questions, aged examples |
| News and time-bound posts | Once, at 12 months | Usually prune or archive, occasionally rewrite as evergreen |
| Reference and documentation | On every product change | Accuracy, not rankings |
Questions people ask
How many pages should one audit cover?
- As many as you can execute on within the same month. Auditing 300 URLs and fixing 12 produces a stale spreadsheet; auditing 25 and fixing 25 produces measurable change and a repeatable cycle.
How long before a refresh shows results?
- Recrawling and reassessment take time and vary by site. Treat anything under a few weeks as noise, and compare year-over-year rather than to last month so seasonality does not read as impact.
Should the publication date be changed after a refresh?
- Only when the substance changed. Google warns against artificially freshening pages, so update the visible date when readers would genuinely benefit from knowing the content was revised.
Is deleting content ever good for SEO?
- Removing pages that have no traffic, no links and no purpose can make a site easier to understand and maintain, but deletion is a last resort after merging and improving have been ruled out. The gain comes from consolidation, not from the deletion itself.
References
- Google Search Console — Performance report — Source for the clicks, impressions, position and query data used in step 1.
- Google Search Central — Creating helpful, reliable, people-first content — Includes the self-assessment questions and the warning about artificially freshening content.
- Google Search Central — Redirects and Google Search — How 301s are handled when merging or pruning pages.
- Google Search Central — Remove a page hosted on your site — The 404/410 and removal options behind the prune decision.
Keep reading
- Why density is a warning light, not a target — Reading repetition as a symptom when you audit a page that reads badly.
- The on-page SEO checklist that is actually checkable — The 21 pass/fail items to run during step 3 of the audit.
- One H1 per page: what actually breaks when you use two — The structural failure most often found on audited pages, and its real consequences.
- The on-page checks a content generator should enforce — The rule set behind the audit, plus a free optimizer for the page you just diagnosed.