What Content Decay Actually Looks Like: 5 Patterns in Search Console

The short version: content decay is a slope, not a cliff. One page loses clicks steadily across weeks or months while the rest of your site holds. In Search Console it shows up as one of five signatures — different combinations of what clicks, impressions and average position did — and which one you're looking at decides the fix. A sharp drop on a single date, across many pages at once, isn't decay; that's something breaking.

Most articles on this describe decay in arrows: "clicks down, impressions down." True, and useless when you're staring at your own numbers, because they never say how far down, over how long, compared to what. So you finish reading and still can't tell whether your page qualifies.

This one gives the numbers. I run this check by hand on my own sites in free Search Console — these are the five shapes I keep finding, the three impostors that fooled me first, and what each one is worth doing about.

The three numbers every pattern is made of

Every pattern below is a different combination of the same three metrics, read together:

• Clicks — people who picked you.
• Impressions — times you were shown.
• Average position — where you ranked, on average, when shown.

(CTR is just clicks ÷ impressions, so it doesn't add a fourth dimension — it's a shorthand for the first two.)

Reading one of them alone is how people misdiagnose. Clicks fell — fine, but the reason is written in what the other two did while it happened.

Here are the thresholds I use to decide whether a page has moved at all. They're my working rules, not industry law, and you should feel free to argue with the exact figures — but pick some, or you'll spend every Monday reacting to a wobble:

• Baseline: the last 4 weeks against the median of the 8 weeks before that. Median, not average, so one newsletter spike doesn't set an impossible bar.
• Decay: recent clicks ≥20% below that baseline.
• Noise floor: the page earns at least ~10 clicks/week at baseline. Below that, a page can "lose 67%" because two people didn't search this week.
• Position slip: average position ≥1.0 worse.
• CTR drop: CTR down ≥15% relative to baseline.

Those are the same numbers RefreshRadar runs on under the hood. Nothing here is a black box — you can reproduce every one of them in the free tool.

Worth saying plainly: the field has converged on roughly this way of splitting the problem — Search Engine Land's four types of content decay draws similar lines. What follows adds the part that's usually missing: the thresholds, where to look, and what to do when a page matches two patterns at once.

Pattern 1 — The slow slide

What it looks like: clicks down, impressions down, average position ≥1.0 worse. All three moving the same way, gradually, over two or three months.

What it means: you're ranking lower than you did. Someone published something better, your page went stale relative to the query, you lost links, or two of your own pages are splitting the same term. This is the pattern people picture when they hear "content decay," and it's the most common one. Which of those actually happened is a separate question with its own test — the six reasons a post loses traffic walks them, and only two of them are fixed by rewriting.

What to do: an actual content refresh — the substance of the page, not the packaging. Update the parts that went out of date, cover what the current top results cover and you don't, and tighten the section that targets the query you're losing. The content refresh checklist walks the specific changes and, just as importantly, what to leave alone.

The honest caveat: a slide of one or two positions inside the top ten is often just SERP churn. Look for a sustained new level, not a bad fortnight.

Pattern 2 — The title problem

What it looks like: impressions flat or up, average position basically unchanged, clicks down, CTR down 15% or more.

What it means: Google is still showing you as often as ever, in the same spot, and people are choosing someone else. That's a click-through problem, not a ranking problem — and it's the cheapest thing on this list to fix, because you're editing two lines of text rather than rewriting an article.

What to do: rewrite the title and meta description so the listing leads with the thing people actually searched for. Full diagnosis, including the >500-impressions / <2%-CTR filter that surfaces these pages in bulk, is in high impressions, low clicks.

The honest caveat: a low CTR at position 9 is just position 9 doing its normal thing. Judge CTR against the position it's earned, never against a flat target — and check Pattern 3 before you blame your own title.

Pattern 3 — The stolen click

What it looks like: average position held or even improved, impressions flat or up, clicks down anyway — sometimes sharply.

What it means: you're winning the ranking and losing the click. An AI Overview, a featured snippet, a People Also Ask block, or a pile of ads is answering the query — or filling the screen — before anyone gets to your result. Ahrefs analyzed around 300,000 keywords in early 2026 and found the top organic result lost roughly 58% of its clicks on queries where an AI Overview appeared.

What to do: first, search the query yourself and look at the page. If an AI Overview is sitting on it, accept that some of that traffic is gone regardless of what you write. Then decide whether the query is still worth serving — often the answer is to shift the page toward the parts of the topic a two-paragraph summary can't satisfy: your own data, a real walkthrough, a decision the reader still has to make.

The honest caveat: this is the one pattern where the correct response is sometimes "nothing." Rewriting a page to win back clicks a machine is intercepting is the most expensive way to learn that lesson.

Pattern 4 — The lost query

What it looks like: the page total is down modestly, but the Queries tab tells a sharper story — one query that used to drive most of the page's clicks has fallen off, while the rest hold steady.

What it means: you didn't lose a page, you lost a term. Two causes dominate. Either another page of yours started ranking for it instead (cannibalization — Google picked the wrong one of your URLs), or the intent behind the query shifted and the results page now rewards a different kind of content than yours.

What to do: in Search Console, filter the Performance report to that one page, open the Queries tab with compare mode on, and sort by clicks difference to see exactly which term left. Then search it. If one of your other URLs is ranking there, you have a cannibalization problem and the fix is to consolidate or differentiate the two pages, not to refresh either of them — here's how to confirm that in Search Console before you act on it. If the results are all a different format than yours — a tool, a comparison, a video — the query moved and matching the page to it is the refresh.

The honest caveat: on a page ranking for hundreds of long-tail terms, a single query leaving is normal turnover. This pattern matters when the departed query was carrying the page.

And one thing that will confuse you the first time: the query rows won't add up to the page total. Search Console omits anonymized queries — terms not searched by more than a few dozen people over a two-to-three month window — from the table entirely, to protect user privacy, and drops them from the totals whenever a query filter is applied. On a small site that can be most of your traffic. So if the Queries tab shows nothing meaningful leaving while the page total is clearly down, don't conclude this pattern doesn't apply — the clicks may simply have left through queries Google won't name. Fall back to average position on the page as a whole: if it slipped, treat it as Pattern 1.

Pattern 5 — The dried-up well

What it looks like: impressions down, average position held or improved, clicks down roughly in proportion to impressions.

What it means: you're ranking as well as you ever did for a thing fewer people are searching for. Demand left. That's not decay in your page — it's decay in the topic, and it's the pattern most likely to get a good page needlessly rewritten.

What to do: check it against Google Trends before touching anything. If interest in the term genuinely fell, leave the page alone. Refreshing it can't create searches. Your options are to consolidate it into a broader page that still earns demand, or to let it sit — and spend the afternoon on a page from Pattern 1 instead, where the traffic is actually recoverable. Deciding that is a skill of its own: when not to refresh a post covers the five verdicts, of which refreshing is only one.

The honest caveat: falling impressions can also mean you dropped out of the long tail — ranking well for your main query while losing dozens of minor ones. Check the Queries tab before you conclude the world stopped caring.

Three things that look like decay and aren't

This is the screen most decay articles skip entirely, and it's the one that saves you the most wasted work.

The cliff. Decay is a slope. If your chart shows a step down on a specific date — flat, then a lower flat — that's not decay, that's an event: an algorithm update, a migration, a robots.txt or noindex mistake, a redirect that broke. The tell is breadth. Decay hits one page while its neighbours hold; an event hits many pages at once. Start with Google's own guide to debugging Search traffic drops, which walks the technical and manual-action causes in order. One important caveat on the cliff-versus-slope rule, though: a core update is not a cliff. Recent rollouts took eleven to eighteen days end to end, so while one is running it draws a slope that looks exactly like decay — telling the two apart properly takes the rollout dates and a specific pair of comparison windows.

The seasonal dip. Compare the same weeks against last year, not against last month. Search Console keeps 16 months of performance data precisely so you can do this. If the curve matches the shape it made a year ago and your position held, the traffic is coming back without you. Decay doesn't repeat on a calendar.

The one-week wobble. A page that gets 12 clicks a week and got 6 last week has not lost 50% of anything. Below roughly 10 clicks/week the percentages are theatre. Filter those pages out before you rank anything, or your list of "worst decaying pages" will be a list of your smallest ones.

When a page shows two patterns at once

It happens more often than the neat taxonomies admit — a page slips two positions and loses CTR, or demand softens and a competitor overtook you. When the signals overlap, work in this order:

1. Position first. If the page lost ranking, that's the dominant problem; a title rewrite won't rescue a page that fell from 3 to 9.
2. Then the cheap fix. A better title costs twenty minutes, so do it alongside the refresh rather than instead of it.
3. Then reassess. Change one substantial thing at a time so the before/after actually tells you something.

If a page matches Pattern 5 plus anything else, treat Pattern 5 as the verdict. There's no fix for absent demand, and everything else you do to that page is decoration.

You found a pattern. Now what?

Naming the shape is the diagnosis, not the treatment. Two things follow.

Rank the pages before you fix any of them. Run the full sweep — how to find declining pages in Search Console — and you'll usually surface more matches than you have weekends. The one worth doing first isn't the biggest percentage drop, it's the one bleeding the most clicks per month: roughly (baseline weekly clicks − recent weekly clicks) × 4.33. That formula, and what to do with the top of the list, is which blog posts to update first.

Then prove it worked. A refresh that you never measure is a story you tell yourself. Record the baseline and the edit date before you touch the page, wait four to eight weeks, and compare the 28 days after against the 28 days before — how to tell if a content refresh actually worked.

Find the pattern, rank by what's at stake, fix the thing the pattern points at, verify. That's the whole loop.

Put a name on your own pages

Doing this by hand is genuinely doable — everything above runs in free Search Console, and I'd rather you learned it than bought something. What's hard is doing it every week, across every page, forever. That's the part that quietly stops happening after the second busy month, and it's why decay wins.

That's the job RefreshRadar does for you: it watches every page in your Search Console, applies the same rules (20% below an 8-week baseline, above the 10-clicks floor), and emails you the ranked list — each page tagged with which of these patterns it matched, in plain English, and the clicks/month at stake.

Connect Search Console read-only and see what's decaying on your site — free — no card, about 30 seconds, and we never post or change anything. You'll see your three worst-decaying pages with the reason on each, so you can check them against the five patterns above on your own data. If nothing on your site is decaying, we'll tell you that instead.