Content decay is usually described as a rankings problem: a page performed well, became outdated and slipped down the results. That definition is too narrow for a useful audit.
A page can lose organic traffic while holding its average position. Demand may have fallen. Search features may be absorbing clicks. A stronger page on the same site may have replaced it. Tracking or template changes can also make healthy content look weaker than it is.
A reliable content decay audit workflow therefore has to separate real deterioration from seasonality, reporting noise and changes in how people discover information. It must then connect each confirmed decline to an appropriate action. Updating every old article is neither a strategy nor a sensible use of editorial resources.
This workflow is designed for business owners, in-house marketing teams and agencies that need an audit they can repeat—not a one-off spreadsheet that produces a long, unprioritised list.
What content decay actually means
I treat content decay as a sustained loss of a page’s ability to satisfy useful search demand or contribute to a defined business outcome. That loss may appear in clicks, impressions, qualified sessions, conversions, citations or assisted journeys.
The word sustained matters. A seven-day dip is not automatically decay. Search demand changes by weekday, season, news cycle and market conditions. Reporting delays and low query volumes can exaggerate small movements.
The word useful matters too. A page losing hundreds of irrelevant impressions may deserve less attention than a commercial guide losing ten high-intent visits.
Common forms of decay include:
- Ranking decay: the page loses visibility for queries it previously served.
- CTR decay: positions remain reasonably stable, but fewer searchers click.
- Demand decay: the topic attracts fewer searches even though the page remains competitive.
- Intent decay: the results page shifts towards a different format, audience or stage of the buying journey.
- Information decay: examples, claims, prices, screenshots, products or recommendations become stale.
- Business-value decay: traffic remains, but conversion contribution or lead quality deteriorates.
- distribution decay: the page becomes less visible through internal links, answer interfaces, newsletters or other discovery paths.
These conditions overlap, but they do not call for the same fix. A title rewrite may help CTR decay. It will not revive a topic whose demand has permanently contracted.
Build the audit around page cohorts, not random URLs
Start with the full set of indexable content pages, then divide it into meaningful cohorts. Comparisons become more credible when similar pages are evaluated together.
Useful cohort fields include:
- Content type: guide, service page, comparison, case study, glossary or news article
- Topic cluster or service line
- Search intent: informational, commercial investigation, transactional or navigational
- Publication and last-substantive-update dates
- Primary owner or subject-matter reviewer
- Organic landing sessions and conversions
- Internal-link depth
- Current indexability and canonical status
Do not rely on the visible “last updated” label. A changed date, corrected typo or automated template edit is not a substantive refresh. If possible, obtain revision history from the CMS or repository.
This inventory is easier to maintain when it is part of an ongoing operating rhythm. The process in How to Build an SEO Operating System for a Small Team provides a broader model for assigning owners, review cadences and decision rights.
The eight-step content decay audit workflow
1. Define the comparison window and minimum evidence threshold
Choose windows that fit the site’s demand pattern and data volume. A common starting point is the most recent 90 days versus the preceding 90 days. For seasonal businesses, compare with the equivalent period in the previous year as well.
No single window is universally correct. High-volume publishers may detect meaningful movement within 28 days. A specialist B2B site with low query volume may need six or twelve months.
Set minimum thresholds before looking at the results. For example, exclude URLs with negligible historical impressions from automated decay flags, then review strategically important low-volume pages separately. This prevents tiny numerical changes from dominating the queue.
Record annotation dates for migrations, redesigns, consent changes, analytics releases, major campaigns and search engine updates. Without those notes, a site-wide measurement break can be misdiagnosed as dozens of page-level failures.
2. Export page and query data from first-party tools
Use Google Search Console to export page-level clicks, impressions, CTR and average position for both periods. Its limitations are important: reported query data may be incomplete, data is aggregated, and average position is not a fixed rank observed by every searcher. Google documents Search Console and search appearance reporting through Google Search Central.
Pull equivalent information from Bing Webmaster Tools where Bing contributes enough activity to be useful. The platform provides search performance and indexing diagnostics; its official entry point is Bing Webmaster Tools.
Join this data to web analytics landing-page sessions, engaged visits, key events and revenue or lead signals where those are implemented responsibly. Keep search-console clicks and analytics sessions as separate measures. They are collected differently and should not be forced to reconcile exactly.
Also collect crawl data covering status code, canonical target, indexability, title, heading, word count, inlinks and crawl depth. A crawler does not prove that a page is useful, but it can quickly reveal technical explanations for apparent decline.
3. Calculate decline using absolute and relative change
Relative percentages alone create misleading priorities. A fall from two clicks to one is a 50% decline; it is rarely the most important issue on the site.
For each URL, calculate:
- Absolute click change
- Percentage click change
- Impression change
- CTR change
- Average-position change, treated as directional rather than exact
- Organic landing-session change
- Conversion or qualified-action change
- Year-on-year change where seasonality warrants it
A simple flag might require a page to exceed a historical impression threshold, lose a specified proportion of clicks and show a meaningful absolute loss. The thresholds should reflect the site rather than a borrowed industry template.
For larger datasets, compare each page with the movement of its cohort or the site as a whole. If nearly every page fell after analytics consent settings changed, individual content updates are unlikely to solve the reported problem.
4. Classify the decay pattern before diagnosing it
| Observed pattern | Likely lines of enquiry | Do not assume |
|---|---|---|
| Impressions and clicks decline; position worsens | Competitor gains, weaker relevance, cannibalisation, technical issue | That adding more words will recover visibility |
| Impressions stable; CTR declines | Snippet quality, changed result features, stronger competing titles | That the body content is the main problem |
| Impressions decline; position remains similar | Lower demand, query-mix change, seasonality | That rankings caused the traffic loss |
| Traffic stable; conversions decline | Intent mismatch, offer changes, broken forms, poorer lead quality | That SEO performance is healthy overall |
| One URL falls while another related URL rises | Consolidation, cannibalisation or intentional replacement | That total topic visibility has declined |
| Search data stable; analytics sessions fall | Tracking, consent, attribution or page-speed implementation | That search demand disappeared |
This classification stage saves considerable editorial effort. It turns a list of losers into a set of testable hypotheses.
5. Diagnose the cause at query, SERP and page level
For every priority URL, review the queries that lost the most clicks and impressions. A page-level total can conceal an important shift: the page may have lost broad informational terms while gaining fewer but more commercially relevant searches.
Then inspect the live results manually. Look for changes in dominant content type, freshness, source profile, media format and search features. Check both desktop and mobile where the audience mix warrants it. Use an appropriate market and avoid treating one personalised result as universal.
Review the page itself against the current task:
- Does the opening answer the likely question directly?
- Are material facts current and attributable?
- Does the page cover the decision criteria users now need?
- Is the author or business qualified to make the claims presented?
- Are examples specific, useful and still valid?
- Can important sections be found through descriptive headings?
- Are internal links helping users move to the next logical step?
- Is a competing page on the same domain serving substantially the same intent?
Do not equate freshness with quality. An evergreen explanation may need no change. Conversely, a recently published page can decay quickly if its premise was shallow or its target intent was misunderstood.
6. Test answer interfaces with a documented methodology
Answer-engine visibility should be tested separately from conventional rankings. I recommend maintaining a fixed prompt set based on real query families, recording the date, interface, market, sign-in state where relevant, exact prompt, response and cited sources.
At minimum, name the interfaces rather than grouping them under “AI search.” A practical test may cover:
- Google Search with AI Overviews: citations and source links may appear in the generated response. Performance is reported within Search Console’s Web search data rather than through a complete, dedicated AI Overview query report, so page-level impact usually has to be inferred alongside broader search data.
- Microsoft Bing Search and Copilot: answers can display linked citations. Bing Webmaster Tools can help with Bing search performance and indexing, but observed answer inclusion and resulting referral traffic should be logged separately rather than treated as a fully reported citation metric.
- ChatGPT Search: responses can provide inline citations and a sources panel. Publishers may observe referred visits in analytics when users click through, but OpenAI does not provide a Search Console-style publisher dashboard exposing every prompt, citation impression or no-click mention. OpenAI’s current developer and platform documentation is available at OpenAI Developers.
These are measurement constraints, not reasons to ignore answer interfaces. The defensible method is to preserve reproducible observations, monitor attributed referrals where available and avoid converting a small manual sample into a market-wide visibility percentage.
For a fuller set of measures, including citations, referral quality and assisted outcomes, see AEO Metrics That Matter Beyond Rankings.
7. Assign one of five actions
Every reviewed URL should finish with a decision. “Needs optimisation” is too vague to enter a production queue.
- Keep: the page remains accurate, useful and aligned with current intent. Monitor it without manufacturing changes.
- Refresh: preserve the URL and core purpose, but update weak sections, evidence, examples, structure, media, title or internal links.
- Reposition: change the page’s target intent or audience because the opportunity has shifted. This is more substantial than a refresh and may require new positioning throughout.
- Consolidate: merge overlapping pages into the strongest destination, update internal links and redirect retired URLs when appropriate.
- Retire: remove content that is obsolete, unsupported, harmful or no longer relevant. Depending on the situation, use a relevant redirect, a clear not-found response or another technically suitable treatment.
A refresh should have a written brief. State what changed in the evidence, what users need now, which sections will be revised, what must be verified and how success will be judged.
8. Prioritise by recoverable value, not traffic loss alone
The largest decline is not automatically the best opportunity. Prioritisation should account for business value, likelihood of recovery, strategic importance and required effort.
A workable scoring model can use four rated inputs:
- Opportunity: historical demand and absolute loss
- Business relevance: connection to qualified actions, services or products
- Confidence: strength of evidence that the cause is understood and fixable
- Effort: editorial, design, development and subject-matter input required
Keep the score transparent. A formula that no stakeholder understands creates false precision. I prefer a simple score accompanied by a one-sentence rationale and an explicit owner.
What a good refresh changes
A substantive refresh is not a mechanical expansion exercise. It improves the page’s fit with the current task while protecting what already works.
Depending on the diagnosis, that may include replacing unsupported claims, adding first-party analysis, clarifying definitions, improving comparison criteria, removing obsolete advice, restructuring the opening, addressing an important sub-question or adding an expert review.
Internal links deserve particular attention. Link from relevant pages that already receive traffic, using anchor text that accurately describes the destination. Also improve onward paths from the refreshed page to useful commercial or educational resources. Where professional support is appropriate, readers can review Akshay Hooda’s SEO, paid media and growth services.
Avoid changing the URL merely to make it look fresh. Stable URLs preserve continuity. Change one only when information architecture, consolidation or persistent misalignment provides a stronger reason.
Use automation for detection, not unreviewed decisions
Automation is valuable for joining exports, calculating period changes, assigning preliminary pattern labels and alerting owners. A scheduled workflow can pull data into a warehouse or spreadsheet, compare rolling periods and create tickets when thresholds are met.
AI can also help cluster lost queries, summarise revision histories or produce a first-pass content diff. Those outputs should remain review aids. Models can overlook commercial nuance, infer causes without evidence or recommend generic additions that make a page longer but less focused.
Keep human approval for consolidation, deletion, factual claims, regulated topics and significant changes in intent. Log the inputs, model version where practical, output and final decision. The audit must remain explainable to editors and clients.
Measure recovery without moving the goalposts
Record a baseline before publishing. Include the comparison window, principal queries, clicks, impressions, CTR, landing sessions, conversions and any manually observed answer citations.
Check implementation immediately, then evaluate performance over an appropriate period. Indexing and demand do not follow an editorial calendar. Early monitoring can identify technical errors, but it is rarely enough to establish durable recovery.
Measure the outcome against the original diagnosis:
- CTR refresh: evaluate CTR while accounting for position and query-mix changes.
- Intent repositioning: examine newly relevant queries and qualified actions.
- Consolidation: assess combined topic-level performance, not only the retired URL.
- Information update: monitor visibility, engagement and conversion contribution.
- Answer optimisation: log citation recurrence and referral quality across the named interfaces.
Not every well-executed refresh will recover. Competitors improve, demand contracts and result layouts change. A useful process records that outcome and decides whether to iterate, consolidate or stop investing.
Concise FAQ
How often should a content decay audit be run?
Quarterly is a practical default for many established sites. High-volume publishers may run monthly detection with quarterly editorial review. Low-volume or strongly seasonal businesses may need longer windows.
How old must a page be before it can decay?
There is no universal age. A page can lose relevance within weeks in a fast-changing topic, while an accurate evergreen resource may remain useful for years. Evaluate performance and validity, not age alone.
Should every declining page be updated?
No. Some declines reflect lower demand, intentional consolidation or irrelevant query loss. Other pages should be retired. Update only when the diagnosis indicates that a realistic improvement is available.
Can AI automate the entire audit?
It can accelerate data preparation, clustering and draft recommendations. It should not independently decide which pages to delete, merge or materially rewrite without human review and business context.
Does changing the publication date help?
Changing a date without materially updating the content does not improve its usefulness. Display dates accurately and make substantive revisions when the topic requires them.
Conclusion: turn decay into a managed editorial decision
A credible content decay audit is not a hunt for old pages. It is a disciplined process for identifying sustained losses, testing competing explanations and choosing the smallest action capable of improving the result.
Start with comparable cohorts and appropriate time windows. Combine search, analytics and crawl evidence. Diagnose changes at query, results-page and page level. Test named answer interfaces with a documented method. Then assign a clear action, owner, baseline and review date.
The strongest outcome is not the highest number of refreshed URLs. It is a smaller, defensible production queue in which every task has evidence behind it—and every completed change can be evaluated against the problem it was intended to solve.