Content decay

Decay is a relative loss, and that changes the fix

A page that decays rarely gets worse. It stays exactly as it was while the SERP around it moves: someone publishes a more complete resource, the dominant intent behind the query shifts, an update re-scores the whole result set, an AI panel answers the question above the fold. That is the entire mechanic, and it explains why the reflex of « rewriting the article » recovers so little. The document is not defective. Its position in a ranked list is a function of every other document on that list, and those changed while yours did not.

The first thing worth separating: decay is not the same thing as content that never worked. Plenty of published pages receive no organic traffic from Google at all. That is a publishing and demand problem, not a decay problem, and if you run a decay audit across an entire library you will burn most of the effort on pages that never had anything to lose. Segment on pages with a genuine click history first, then look at the slope. A library of 4 000 URLs usually yields 150 to 300 legitimate decay candidates, and the rest belongs in a different conversation about whether it should have been published.

The slope matters more than the drop. A loss concentrated on a single date is an update, a technical event, or a SERP feature appearing. Decay is gradual and boring: a 2026 compilation of refresh studies puts the average rate at roughly 1.21 % of traffic lost per week, and reports that unmaintained decaying pages typically shed 20 to 30 % of their organic clicks over six months. Nobody sees that in a monthly report. Everybody sees it a year later, when a category is down by a third and no individual page looks guilty enough to explain it.

This is also why decay is a maintenance discipline rather than an editorial one. The question is never « is this article good ». It is « does this article still hold the position it earned against what is on the page today », which you answer by looking at the current top ten, not at your own draft.

What actually changed by 2026

Three shifts turned decay from a slow background tax into the dominant traffic story of the last two years, and only one of them is about content quality.

The first is displacement by AI answers. Semrush, analysing more than 10 million keywords, found AI Overviews on 6.49 % of queries in January 2025 and 13.14 % by March 2025, a 102 % relative increase in two months, with about 88.1 % of the triggering queries being informational. Informational is exactly where aging evergreen content lives. Seer Interactive, working from 25.1 million impressions, measured a 61 % fall in organic CTR on informational queries carrying an AI Overview between June 2024 and September 2025. Zero-click behaviour tracks the same curve: SparkToro and Datos put roughly 64 % of Google searches ending without a click, Bain and Dynata measured about 83 % when an AI Overview is present in December 2024, and Semrush reported about 93 % inside AI Mode in September 2025. The operational consequence is a failure mode that did not exist in 2020: your position holds, your impressions hold, and your clicks fall anyway.

The second is update cadence and violence. Google ran four core updates in 2024 and three in 2025. The March 2026 core update touched over half of monitored sites, with declines of 20 to 35 % common and some properties losing more than half the traffic on their strongest pages, at Semrush Sensor volatility of 9.4 to 9.5 out of 10. Every one of these events re-scores quality natively, and a page that was merely adequate does not survive three of them in a row.

The third is freshness itself, and here I would push back on the consensus. A 2026 synthesis of 15 ranking-factor studies ranks content freshness as the sixth Google ranking factor with an estimated 6 % weighting, behind relevance, backlinks, E-E-A-T, technical SEO and user engagement. The same source reports that updated posts see on average 106 % more organic traffic than non-updated ones, aggregating over 500 data points. Read those two numbers together rather than separately: 6 % weighting means freshness alone moves almost nothing, and the +106 % figure is heavily selection-biased because teams refresh the pages that still had something worth defending. The refresh works because of what was added, not because the date moved.

Intent drift deserves its own line because it is invisible in any content audit. The canonical example doing the rounds in 2026 guides is « LLM »: a query that meant Master of Laws and now means large language model. Nothing on the legacy page broke. The query simply stopped pointing at it. Checking the live SERP for your top queries once a quarter catches this. Reading your own article never does.

Measuring decay without fooling yourself

Search Console is enough for the diagnosis, provided you decompose rather than look at a single traffic line. Three signatures cover most of what you will find. Position stable, impressions stable, clicks down means displacement: a SERP feature, an AI Overview, or a competitor stealing the snippet. Position down with impressions down is a genuine competitive or quality loss and a real refresh candidate. Impressions up with position down and clicks flat means your query set is drifting toward adjacent, less qualified terms, which is usually a symptom of a page that has been broadened once too often.

The comparison window is where most audits go wrong. Compare 28 days against the same 28 days of the previous year, not against the previous month, otherwise seasonality writes your conclusions for you. Set a floor of around 100 clicks in the baseline period, below which the noise exceeds the signal. And run the analysis at cluster level rather than URL level, because consolidation, canonical shifts and internal cannibalisation move clicks between your own URLs without anything actually decaying. A page can look like it lost half its traffic when a sibling absorbed it, which is a reporting artefact, not a problem.

The competitor half of the diagnosis needs a rank tracker or a tool like Ahrefs or Semrush, because Search Console cannot tell you who took the position. What you want is the identity of the pages that gained on your losing queries and what they contain that you do not. That is a gap analysis on a query set you already own, which is far cheaper and more reliable than a gap analysis on keywords you have never ranked for.

One threshold we apply in practice: if a page has lost more than 40 % of its clicks year over year and still sits in the top twenty, it is worth real work, because the ceiling is provably reachable. Below position thirty, a refresh is usually a rewrite in disguise and should be budgeted as one.

Where decay hits a netlinking operation

This is the part the ranking guides skip. Decay is not only a content problem, it is a link-value problem on both sides of the transaction.

As a buyer, the article hosting your link decays like any other article on that site. The link itself does not disappear, and it keeps passing whatever it passes, but the document it lives on progressively loses impressions, loses internal prominence as newer content pushes it down the archive, and gets crawled less often. A placement bought on a page that will shed most of its visibility within eighteen months is a depreciating asset. The benchmark to hold in your head comes from the HubSpot Content Performance Study 2024, compiled by Starr Conspiracy in its 2025 B2B benchmarks: 67 % of B2B blog posts lose at least half their organic traffic within 18 months, with technical content decaying at a median of roughly 14 months and higher-level strategic content lasting around 24. Unless the publisher maintains its archive, the half-life of your placement is the half-life of their average post.

Which turns into a selection criterion nobody applies: before buying, look at whether the host still ranks with content it published two or three years ago. That single check separates a media property from a content dump far better than any authority score. It is also why placing an editorial piece that is still earning after its first quarter costs more to produce than a placement on an abandoned archive, and should. Across the 50 owned French editorial media we run in-house, maintenance of the older archive is treated as part of the product rather than as an optional refresh project, because a link sold in January should still sit on a page that matters in December. If you want to judge that for yourself, the useful exercise is to open a catalogue you can filter without creating an account and check the publication dates of the pages that still hold traffic.

The publisher side has the mirror obligation. Decayed pages are also where internal link equity gets stranded: a hub that lost its rankings still distributes links, but it distributes them from a document search engines revisit rarely. Refreshing the pages that carry the most internal links out is usually higher leverage than refreshing the pages that lost the most traffic, and it is one of the few interventions that compounds across a whole cluster built to own a subject rather than helping one URL. When acquisition is planned quarter by quarter rather than in bursts, that maintenance rhythm and the link rhythm can be calibrated against the same content calendar, which is the only version of this that survives a core update.

What we see go wrong in decay audits

Date stamping without substance is the most common and the most useless. Changing the modified date, swapping « 2025 » for « 2026 » in the title and republishing moves a 6 % weighting factor by a fraction of itself. Dates are supposed to reflect real changes, and the practice mainly succeeds at destroying your ability to measure whether real refreshes work, because every page now looks recently updated.

Prioritising by absolute loss instead of by recoverable ceiling comes second. The page that lost 3 000 clicks from position 4 to position 9 is worth attacking. The page that lost 3 000 clicks because the query now triggers an AI Overview is not, at least not through a rewrite: that one needs a different query target or a format the answer panel cannot summarise, typically original data, tooling, or a transactional angle.

Pruning too aggressively is the expensive mistake. A page with zero clicks can still hold referring domains and sit on an internal path that matters. Deleting it converts an asset into a 404 and a lost signal. Redirect into the page that genuinely absorbs the intent, verify that target ranks before firing the redirect, and keep the mapping documented. Consolidation has the same trap in reverse: merging two pages that answer two distinct intents produces one page that answers neither, and the merged URL usually ends up below where the better of the two sat.

The blind spot almost everyone shares is non-editorial decay. Category pages, product pages and commercial landing pages decay too, and it is harder to see because branded and navigational queries prop up the aggregate while the non-branded head terms slide underneath. Split branded from non-branded in Search Console before concluding a commercial page is healthy. Finally, the audit that runs quarterly, produces a spreadsheet and generates no work order is not measurement, it is theatre. Attach a decision rule to each signature before you run it: refresh, retarget, consolidate, redirect, or leave alone. A decay report without a default action gets read once and archived.