A page is published, starts ranking within a few months and brings in steady traffic for a year. Two years later, the same page's performance chart still has a line on it, but now it points down. There is no single day you can point to and say "this is where it broke." The site is up, the page is indexed, and not a single sentence has been removed from the text. Yet half the clicks it had six months ago are gone.
This pattern is called content decay. The word "decay" is a little misleading, because it implies the page is breaking down on its own: in most cases the page does not change, its surroundings do. The competing answers to the same question get better, the way the question is asked shifts, and the layout of the results page is rebuilt. The page falls behind because it stands still.
That is why decay is not a malfunction, and treating it like one leads to the wrong fix. The right decision depends not on the size of the loss but on its cause: the same sixty percent drop in clicks can call for an update on one page, a merge on another and no action at all on a third. Diagnosis is where these three paths split.
A sudden drop and a slow slope are not the same thing
The shape of the chart gives you the first half of the diagnosis. Google's own guide to search traffic drops does not tie a decline to a single phenomenon; it defines several separate causes: algorithmic updates, spam or security-related actions, seasonality and changing user interest, the site's own technical issues, and reporting glitches. There is no item called "content decay" on that list, because decay is not a cause but a shape. Some items on the list produce a steep cliff, others a long slope.
The practical distinction is this. A steep drop that happens within a few days is almost never decay: the first suspects are the page dropping out of the index, a redirect error, server availability, an accidentally added blocking tag or a broad algorithmic update. A slope that stretches over months, with no break point on any single day, points to one of two candidates: decay or seasonality. The only reliable way to tell them apart is to look at the same period last year, because a seasonal drop repeats in the same months every year.
Why does content lose value on its own?
Decay does not have a single engine. It has five mechanisms that work independently of one another and often overlap on the same chart.
Competitors getting better. A page does not earn its ranking with an absolute score but by comparison with the alternatives available at that moment. Google says this explicitly in its traffic drop guide: if there is more deserving content, its systems will keep ranking it well. So you can slip without anything happening on your page, simply because someone published a page that answers the same question better. How the crawling, indexing and ranking stages of search work is covered in detail in What Is SEO; the key point here is that ranking is a selection, not a reward.
Freshness expectations. Freshness does not carry the same weight for every query. Google's explanation of its search signals gives a direct example: when someone searches for a current news topic, content freshness plays a much bigger role than it does for a query looking for a dictionary definition. The takeaway is that the oft-repeated industry line "content loses its freshness after a certain time" is not a query-independent rule. Freshness is not a calendar; it is a characteristic of the query.
Search intent shifting. The words of a query can stay the same while the expectation behind them changes. When a topic is new, people look for definitions; as it spreads, they look for implementation; as it matures, they look for comparisons and prices. Because the page was written for the original intent, it keeps giving the right answer, just not to the question being asked anymore. The clearest indicator of this is the results page itself: if the type of pages holding the top spots for the same query has changed, the intent has shifted.
The results page being rearranged. What sits above the organic results changes over time. When a new results block, a featured answer area or an AI summary is added, your page's chances of being clicked drop even if it holds the same position. In this case impressions hold steady while clicks melt away. On the chart it looks like decay, but it has nothing to do with the page's content.
The page's own technical decline. A post that gets links from the menu, homepage and category listings on the day it is published quietly loses those links as it slides back through the archive. Template changes, speed regressions, chained redirects and internal links that break over time all push in the same direction. This item is skipped in most decay analyses, yet it is the only one you can fix without touching the content at all. Running a page-level technical check with an on-page SEO analysis and ruling this out before you make a content decision saves time.
Not every page decays at the same speed
There is no universal number for how fast content decays. Thresholds like "content decays after this many months" are averages for a particular site, sector and period; they are not your page's schedule. Using such a number as a rule means spending refresh effort on pages whose calendar is up rather than on pages that are actually decaying.
What determines the speed is not the page's age but the character of the query it serves. Queries involving prices, version numbers, regulations, quotas or dates, or describing a tool's interface, go stale quickly: the answer itself changes, so the page becomes wrong. Concept definitions, process explanations and why-and-how questions go stale slowly; the answer does not become wrong, an alternative that explains it better simply appears. Decay in the first group is an accuracy problem and is urgent. Decay in the second group is a competition problem and can be prioritized.
How to look for decay in Search Console
Detection starts not by looking at pages one by one but by comparing periods. The compare tab of the date filter in the Performance report answers two questions separately: comparing the last three months with the previous period shows what has changed recently, and comparing the last three months with the same period last year shows whether the change is seasonal. Seasonality is the most common false positive, and only the year-over-year comparison rules it out.
With the comparison on, switch to the pages tab and sort the decline not by percentage but by the absolute size of the click difference. Sorting by percentage moves a page that went from twenty clicks to five to the top of the list and hides a page that went from four thousand clicks to three thousand. Once the candidate pages are identified, drill down to the query breakdown for each: whether the loss is spread across all queries or comes from a single query narrows the diagnosis considerably. What the report's metrics really measure, and especially why average position is not a ranking, is covered separately in our Search Console data tracking article.
There is a structural limit here, and it kicks in exactly where it hurts most in decay analysis: the Performance report keeps sixteen months of history. For a page that peaked two years ago, the peak itself is no longer in the report; you only see part of the decline that followed. In other words, the reference point you need most is older than the data you have. The way around this limit is not to bring back history but to start storing it from today: performance data can be kept on your side through the Search Analytics API or bulk data export. If this has not been set up, the lost history does not come back and your decay assessment is limited to the sixteen months you have.
What to read when impressions and position diverge
A drop in clicks on its own does not give you a diagnosis. The meaningful signal lies in how clicks, impressions and average position move together, and there are four typical divergence patterns.
If impressions fall while position holds steady, the page has not lost its spot; the query has shrunk. Either fewer people are searching for it, or the query itself has shifted to a different phrasing. This is the only pattern unrelated to content quality, and often the right decision is not to intervene. If impressions fall along with position, the picture flips: the page has been pushed back in the candidate pool, which means you are looking at competitive displacement. If only clicks melt away while position and impressions hold steady, the problem is how the page appears on the results page; a new element may have appeared above you, or the title and description may no longer match the query. In the fourth and most misleading pattern, average position rises while clicks fall: the page has dropped out of strong queries and started appearing for weak, narrow ones, and the average looks good because the mix of queries behind it has changed.
Decision table: from symptom to intervention
| Symptom on the chart | Likely diagnosis | Verification step | Intervention |
|---|---|---|---|
| Position steady, impressions falling | Query demand dropped or phrasing shifted | Compare the query breakdown across two periods, run a year-over-year comparison | If demand really dropped, do nothing; if phrasing shifted, bring the new phrasing into the content |
| Position and impressions falling together | Competitive displacement | Search the query and see what the top-ranking pages offer that you don't | An update or deeper treatment that covers the missing information need |
| Position and impressions steady, clicks falling | Weaker appearance on the results page | Check what sits above the organic result on the query's results page | Refresh the title and description for the query; a presentation fix, not a content fix |
| Average position rising, clicks falling | Query mix has weakened | Put the two periods' query lists side by side and find the strong queries that disappeared | A rewrite targeting the main query you lost |
| Several of your pages rotate for the same query | Your own pages are replacing each other | With the query filter on, list every page that appears for that query | Merge into the strongest page and 301 redirect the others |
| The drop repeats in the same months every year | Seasonality | Compare the last three months with the same period last year | Do nothing; plan around the calendar knowingly |
| Steep drop within a few days | Technical issue or algorithmic change | Check the page's index status, availability and blocking tags | Technical diagnosis first; don't assume decay and start writing content |
Four intervention options and how to choose
Once the diagnosis is clear, you have four options on the table, and they are not lighter or heavier versions of each other but answers to different problems.
Updating applies when the page answers the right question with the right structure but some parts have gone stale. The outdated section changes, the rest of the skeleton stays. It is the lowest-cost option and the natural response to decay caused by outdated information.
Rewriting is needed when the query has stayed the same but the expected answer has changed. Replacing parts does not help here, because the problem is not in individual sentences but in how the page is built. The URL is kept and the body is rebuilt from scratch. This is the option for intent shifts and for pages that no longer hold up against competitors.
Merging with a 301 redirect comes into play when several weak pages meet the same information need with different wording. These pages eat into each other's rankings, and none of them can get strong enough on its own. The most comprehensive one is chosen as the target, the usable sections of the others are moved there, and the old addresses are redirected to the target. In practice, this is the hardest decision to reverse, so settle on the target page before you set up any redirects.
Deliberate retirement is the right decision when the query itself has disappeared, the topic is no longer within the scope of the business, or the page is attracting the wrong audience. Retirement does not have to mean deletion: the content can be consolidated into another page, or the page can be removed from the index while remaining accessible. Retiring a page is not a failure but a decision that frees up resources.
The fifth, often skipped option is doing nothing. Refreshing a page for a query where demand has genuinely shrunk does not bring the loss back; it just burns effort. Every page that goes on the intervention list should first pass the question "can this loss be recovered?"
Why changing the date doesn't count as an update
The most common shortcut applied to decaying pages is pushing the visible date forward without touching the content. There is a technical reason this doesn't work: Google does not read date information from a single field. As its own documentation states, because any single signal can be wrong, its systems look at multiple factors to produce a best estimate of when a page was published or meaningfully updated. The visible date is only one of those factors.
Seen this way, the date is not a lever but the output of a real update. When content changes meaningfully, changing the date is correct and honest with the reader. Changing only the date while the content stays the same produces two things: no ranking gain, and the loss of your own record of when that page was last genuinely touched. The second loss is more expensive in the long run, because at the next audit you cannot tell which pages are actually current.
Which page should you tackle first?
The decay list is always longer than your capacity to work through it, so the real work is prioritization. The habit that distorts prioritization most is ordering pages by what percentage they lost. Priority comes from reading four criteria together.
- The absolute size of the loss. Not the percentage, but how many clicks are gone. Large percentages on small pages needlessly crowd the top of the list.
- The page's commercial proximity. A loss of a hundred clicks on a page close to a service or product decision can be more costly than a loss of five hundred clicks on an informational page.
- Distance to the threshold. Pages sitting near the bottom of page one or just past it come back with a reasonable improvement. For a page that has fallen far back, an update is rarely enough; the decision there is usually a rewrite or a merge.
- Whether the diagnosis can be fixed. A loss caused by shrinking demand will not come back no matter the effort. These pages should be removed from the list, not moved down it.
Turning these four criteria into a scoring formula usually breaks down, because the weights change by site and period. Using the criteria as a sequential filter gives sturdier results: first remove what cannot be fixed, group the rest by commercial proximity, and within each group sort by absolute loss.
When to measure after a refresh
The most common post-intervention mistake is checking the results within a few days and panicking into a second change. Google's own guide gives a clear frame for expectations here: some changes can take effect within a few days, others can take months, and in general it is reasonable to wait a few weeks before assessing results. It also repeats one caveat often: there is no guarantee that your changes will have a visible effect on search results, because if there is more deserving content, its systems will keep ranking it well.
A second fact follows from this. Some improvements only become visible at the next core algorithm update. Since Google makes smaller updates continuously and without announcement, this wait is not always tied to a big update calendar, but accepting that it can take months changes your measurement plan.
Three habits make measurement reliable. Record the date of the intervention on your side, because that date defines the comparison window. Look not at the site total but at the queries you targeted on the page you targeted; site-wide numbers get polluted by the movement of other pages. And do not make a second change to the same page before the measurement window closes, otherwise you cannot tell which intervention produced what. When you are undecided, changing a single variable on a single page produces more information than fixing five things at once.
Frequently Asked Questions
Does deleting a decaying page hurt the site?
The decision to delete is not harmful in itself; how the URL is closed is what matters. If the page is about a topic with genuinely no demand left and cannot be redirected anywhere, closing it is normal. But if the page receives external links or still gets impressions, redirecting it to a related page instead of deleting it outright preserves the value it has built up. If there is no suitable page to redirect to, removing it from the index is a less costly intermediate step than deletion.
Should the URL of an updated page change?
No, keep the URL unless you have a compelling reason to change it. The links and past performance a page has accumulated are tied to its address; changing the address trades the gains of the refresh for an unnecessary redirect risk. A URL change only comes up when the topic has genuinely turned into something else and the current address misdescribes the new content.
How often should I update the same post?
There is no fixed interval, and calendar-driven updates often end up cosmetic. Update frequency is set by how quickly the query the page serves goes stale: for topics where the answer can change, update when a change happens; for conceptual topics, update when performance data shows a decline. The trigger is not the calendar but either the facts changing or the chart starting to slope.
I updated the page but the decline continues. Should I roll back?
Rolling back before the measurement window is up destroys the only data you have. First check whether the update actually matched the diagnosis: if you applied an information refresh to a page diagnosed with competitive displacement, a continuing decline is the expected result, and the mistake lies in the diagnosis, not in the refresh itself. A rollback only makes sense if the page's scope was narrowed during the update, that is, if an information need the page used to answer was removed from the text.
