What Content Decay Is, and Four Things That Look Exactly Like It

Content decay is the slow loss of search visibility on a page nobody touched. It is a slope, not a cliff: traffic and rankings drift down over months while the post itself sits there unchanged.

You may also have met the phrase in a different argument, the one about online services getting worse as they grow. That is a separate idea with its own name, and it is not what this page is about.

What follows is the part almost nobody writes down. Four other things produce exactly the same falling curve, two of them get fixed by doing nothing, and one of them gets worse if you do the thing every article on this topic tells you to do.

What content decay is, and what it is not

Content decay is the gradual decline of a published page's organic traffic and rankings over time, with no single event to point at. The page was not edited. Nothing broke. It simply stopped holding the ground it used to hold.

The word that matters in that definition is gradual. A decayed page loses ground over months or years, in a line you only notice when you widen the date range. That is the whole discriminator, and it rules out the two things readers most often confuse it with.

It is not a penalty. Nothing was sanctioned, no message arrived in Search Console, and there is nothing to appeal. Manual actions are rare, they are announced, and they hit all at once.

It is not an algorithm update either. Updates land on a date. They hit a whole class of site at the same time, and your competitors move with you. If your chart bends over eight months while the rest of your category holds, no update did that.

Why a page you never touched starts losing ground

Three causes explain most of it, and every serious guide on the subject names the same three.

The query changed meaning. What people expect behind those words is not what they expected the year you published, and your post still answers the old question. Nothing about your page got worse. The target moved.

Someone published fresher. Not always better, only more current. Freshness is worth more now than it used to be, and not only in classic results: Ahrefs reports that URLs cited by AI assistants are on average 25.7% fresher than organic results, though the sample behind that figure is not published. If you want the longer version of what AI-generated rewrites do to rankings, we wrote it up in will AI rewriting hurt rankings.

The facts went stale. Prices, version numbers, screenshots, the year in your opening line. Readers notice those faster than any algorithm does, and they leave.

Those three are real, and they are also where every other article stops. The problem is that they explain a minority of the curves that go down. The rest of this page is about the majority.

Losing impressions is not the same as losing rank

A page can lose half its traffic without losing a single position. Most guides treat clicks, impressions and rank as one blob called traffic, so they never separate the cases, and the reader ends up rewriting a page that was never the problem.

Two numbers, read together, sort it out. Here is the whole grid:

The middle row is the expensive one, because it looks identical to the first from a traffic report. You are still in position 4. Fewer people are searching, or the results page above you now answers the question without a click. An AI panel that summarizes your answer costs you clicks at a rank you never lost.

Rewriting that page will not bring those impressions back. Nothing you do to the page will, because the page is not what changed.

One practical consequence: impressions move before clicks do, so impressions are the early signal. Search Console shows both, side by side, for any date range you choose. Watch the impression line on a twelve month window and you will see the bend well before the traffic report makes it obvious.

Four things that look exactly like content decay

Each of these draws the same falling line. Each takes about a minute to rule out.

1. Seasonality. Your category has a season and you are out of it. Compare the same weeks last year rather than the previous quarter. If the shape repeats, you found your answer.

2. A gap in your own measurement. A property that changed hands, a tag that stopped firing, a consent banner that started blocking. This one deserves a closer look than the others, because the conclusion is the opposite of what the chart suggests. Decay is a slope. A collection failure is a wall: the line does not bend, it steps. Find the date of the step and check what you changed that week. And know that the hole never fills in later, because nobody was collecting.

3. The page fell out of the index. Not one of the seven competing guides we read for this page mentions it, and it produces an identical curve. Search Console's URL inspection answers it in one query. A noindex tag someone added, a canonical pointing elsewhere, a robots rule that grew a new line.

4. Another one of your own pages. That one has its own section, because the fix for it is the reverse of the fix for decay.

When the competitor is your own second article

Two of your posts can split one query between them, and the result looks exactly like decay. Most guides list this as a cause of decay. It is not a cause, it is a competing diagnosis, and that distinction decides what you should do next.

Here is why it matters: refreshing a cannibalized page makes the collision worse. You are strengthening one of two pages that are already dividing the same signal, and you are usually strengthening the wrong one, because the page you noticed is the page that lost.

Telling the two apart by eye does not work, which is why Unstale does it by measurement. Each post gets a lexical fingerprint of its 40 rarest terms, and two posts are paired when their overlap scores above 0.40 against the shorter of the two fingerprints. In practice that means at least 8 rare terms in common before anything is reported. The free plugin compares the 2,000 most recent posts on the site.

Those numbers are not guesses. We calibrated them on September 6, 2026 against 738 real pages drawn from 62 sites, including articles, FAQ blocks carved out of them, and thin pages like contact and legal notices. Across 14,469 pairs of short, unrelated pages, that floor paired none of them. That is the test worth asking any decay tool to pass, and as far as we can tell from reading the field, nobody else publishes the answer.

What this criterion does not do is tell you which of the two to keep. That decision, and the merge that follows it, belongs to refresh old WordPress posts, and choosing between merging and cutting is its own question: prune or refresh.

The 20% rule, and where it comes from

If you have read about content decay before, you have met a threshold: flag any page that lost 20% or more year over year. It is a reasonable number. It is also unsourced.

Two of the most cited pages on this query publish it. Ahrefs gives it as an audit rule in a page dated March 13, 2026. SEO.com gives the same figure in a page dated September 4, 2026. We read both on September 16, 2026. Neither names a study, a sample, or a measurement. It is the one decision rule a reader carries away from the whole category, and nobody can defend it.

The deeper problem is not the number, it is what the number requires. A slope needs a series: at least two comparable periods, a consistent property, and a tool holding the history. If your measurement has a gap in it, and by the time you go looking it usually does, the slope you compute is fiction.

Age needs none of that. It reads off the post itself. That is why the free Unstale plugin sorts by age rather than by slope, with a default threshold of 180 days and three bands derived from whatever threshold you set rather than from fixed dates. Drop the threshold to 90 and the bands move with it.

Be clear about what that buys you: age does not measure decline. It measures elapsed time. It is a place to start looking, not a diagnosis, and the four sections above are what turns one into the other. If you want the sorting part in detail, that is find outdated content on your site.

Before you refresh: what it costs, and what it can cost you

Suppose you have ruled out all four look-alikes and the page really is decaying. Two facts before you start, and neither appears in any of the seven guides we read.

A refresh has a price, and a backlog has a big one. Every one of those guides recommends refreshing. Not one of them says what it costs. If you are rewriting with a model, the bill is measured in bytes sent and returned, plus the hours someone spends reading the result before it goes live. On a backlog of a few hundred posts that is a real number, and it is worth having before you commit: what an AI refresh costs.

A refresh can also make it worse. A rewrite is a reversible bet, not a guaranteed gain. Pages do come back lower. That risk is manageable, but only if you plan for it up front: read the change before it publishes, and keep the ability to roll it back. What actually happens to rankings after an AI-assisted rewrite is will AI rewriting hurt rankings, and republishing without losing ground is republish without losing rankings.

Questions people ask about content decay

Is content decay a Google penalty?

No. Nothing was sanctioned and there is nothing to appeal. A penalty arrives on a date and is announced in Search Console; decay is a slope that takes months and announces nothing.

How long before a page starts to decay?

There is no age at which it starts. There is a window at which you can see it: you need at least two comparable periods of data, which is why most of the field works year over year. Age is a useful way to decide what to look at first, not a measure of decline.

Does changing the publish date fix content decay?

Changing the timestamp alone changes nothing about what actually went stale. Whether the visible date should move at all when you do rewrite the content is a separate question, and it has its own page: changing the publish date.

Do all pages decay, or only old ones?

Age is correlated, not causal. A page declines when the query or the competition moves, and older pages have simply had more time for both to happen. A page published last month can decay if its query shifts next week.

How long should I wait before judging a refresh?

Long enough for the page to be recrawled and served again, and no less than that. None of the guides we read give a number for this, and we are not going to invent one. What we can say is that judging a refresh on the week after you publish it tells you nothing.