Content Decay: Find Declining Pages and Plan an Evidence-Based Refresh

Compare complete periods before labelling a page as decaying

Choose two equal-length periods that contain the same weekdays. For a page with steady demand, two complete 28-day windows can be a useful starting point. Examine a longer trend and the equivalent period last year when seasonal demand is plausible and data is available.

Keep the Search Console property, search type, country and device filters consistent. Record the extraction date and leave incomplete recent days out. The connector’s freshness date is a conservative estimate, so it should not be treated as a guarantee that every recent day is final.

Calculate the absolute click loss and percentage change for each comparable page. A fall from 1,200 to 720 clicks is a loss of 480 clicks and a 40% decline. A fall from five to three is also 40%, but it represents two clicks and needs a different priority.

Google’s traffic-drop guidance recommends examining affected pages and queries, seasonality and wider demand. A site migration, technical fault or market-wide decline needs its own investigation before you rewrite an otherwise useful guide.

Separate click loss, impression loss and changing demand

Collect page clicks, impressions, CTR and average position for both periods. Start with page-level rows, then inspect queries for the largest losses. Keep original totals beside any calculated differences, with the filters and aggregation recorded.

When impressions hold steady and clicks fall, examine the search results, title, snippet and competing answers. A changed result layout can affect CTR. A similar average position across the whole page can also hide different query-level movements.

When impressions and clicks both fall, inspect the same query groups over time. Check demand separately, for example in Google Trends, and compare similar pages. Search Console alone does not establish that fewer impressions resulted from outdated content.

A page missing from one limited response has no observed row in that result. Narrow the request to that exact page before calling it a zero. The connector supports up to three dimensions and 1,000 rows per request, and the underlying API returns top rows rather than a guaranteed full inventory.

Prioritise a refresh queue using absolute losses and page purpose

This hypothetical comparison uses two complete 28-day windows with identical filters. CTR is clicks divided by impressions. The figures illustrate different investigations, without claiming customer results.

The setup guide has lost 480 clicks, the largest absolute loss here. Its stable impression total supports a closer look at CTR and query mix. That evidence still leaves several possible causes, including a changed results page and a title that no longer describes the answer.

At unchanged 30,000 impressions, a planning scenario of 3% CTR would produce 900 clicks, or 180 more than the recent period. Treat that as a scenario for deciding whether an investigation is worthwhile. It is not a prediction that a refresh will recover those clicks.

Add the page’s business purpose and the effort required to correct it. Your website content audit can supply ownership, inventory and website outcome evidence. Keep search clicks separate from qualified leads or sales.

Inspect the actual answer before deciding what to update

Read the page alongside the queries that lost clicks. Check screenshots, product names, instructions, examples, cited sources and the next action offered to the reader. Write down the exact statement or missing step that needs correction.

SourcetoMCP retrieves connected account data for this analysis. Its Search Console report does not read your page’s prose, crawl the whole site or judge every competing answer. Supply the current text or inspect the page yourself before accepting an AI recommendation.

Check indexing and canonical evidence when the decline follows a migration or URL change. If several pages answer the same question, use the keyword-cannibalisation workflow to inspect their purposes before proposing a merge.

Google’s helpful-content review questions support a factual review of completeness and usefulness. Correct a broken instruction, add a missing worked example or remove an unsupported claim. Changing a date alone does not correct the underlying content.

Ask AI for a refresh brief with evidence and unresolved checks

Review content-decay candidates for [property] across [two complete periods]. Make no website or file changes. Use Search Console web search with [country/device filters]. Rank observed pages by absolute click loss and show clicks, impressions, CTR and position in both periods. Investigate selected pages with page-and-query reports, retaining limits and warnings. Separate demand, technical, query-mix and content hypotheses. Use the supplied page text to propose exact corrections. Return each URL, evidence, missing checks, proposed edit, owner and follow-up period. Do not invent missing rows, forecast recovered traffic or recommend automatic deletion.

Ask for five well-supported candidates before commissioning a site-wide rewrite. A compact queue makes it easier to verify each claim and supply the missing page text. Save the approved scope with the original response and the filters used.

The SEO quick-wins guide covers smaller changes on existing pages. Use this decay investigation to decide whether a page needs that small correction, a deeper review or no editorial change yet.

Measure the approved update against a documented baseline

Record the substantive edits, affected URLs and publication date. Keep a copy of the original content and choose a follow-up window suited to the page’s traffic. Avoid declaring success from one unusually strong day.

Compare the same page and query groups using the original filters. Track visibility and clicks, then assess website outcomes separately. Note other changes, campaigns and demand shifts during the comparison so they remain visible when interpreting the result.

If the evidence remains small or contradictory, keep the result inconclusive. The next action should name the additional period or specific inspection needed before you spend more time rewriting the page.