Why Your Best Blog Posts Lose Traffic (and Which Cause Is Yours)

The short version: a blog post loses organic traffic for one of six reasons — someone overtook you, the query changed shape, the results page started answering it, you did it to yourself, your links died, or demand left. Only the first two are fixed by rewriting the page. Age is not on the list.

Every article on this subject hands you a list roughly like that one and stops. Which is where the actual problem starts, because you're not short of possible explanations — you're standing in front of one page that used to get 400 clicks a month and now gets 140, and you need to know which explanation is yours. So each cause below comes with a test that rules it in or out, and an honest note on what it costs you to guess wrong.

Your post didn't get old. Something happened to it.

The standard opening line is that content goes stale. It's worth checking that against the data, because the data says something close to the opposite.

Ahrefs studied 1.3 million random keywords and found that 72.9% of pages in Google's top ten are more than three years old, that the average #1 ranking page is about five years old, and that only 13.7% of top-ten pages were under a year old. Old pages aren't losing. Old pages are what page one is mostly made of.

So "it got old" can't be the cause. Age is the normal condition of a page that ranks. Stale is a word we apply after the fact to a page that already lost, which makes it a description of the outcome rather than an explanation of it. Something specific changed — and there are only about six things it can be.

First, rule out the two things that aren't causes at all

Before you diagnose decay, make sure you're looking at decay.

The event. If your chart shows a step down on one date — flat, then a lower flat — that isn't decay, that's something breaking or something changing all at once. The tell is breadth: decay hits one page while its neighbours hold, an event hits a lot of pages together. Google's guide to debugging Search traffic drops walks the real candidates in order — algorithmic updates, technical faults, security issues, spam violations, site moves, even reporting glitches. Work through that first. But don't lean on the step-down shape too hard: a core update takes eleven to eighteen days to roll out, so while it's running it makes a slope, not a step, and it can move one topic cluster while the rest of your site holds. Content decay or an algorithm update? has the check that actually separates them.

The wobble, and the season. A page that earns 12 clicks a week and got 6 last week has not lost half of anything. Below roughly 10 clicks a week the percentages are theatre — that's the floor I use before a page is allowed onto any list. And before you call a real drop decay, compare the same weeks against last year, not against last month; Search Console keeps 16 months of data precisely so you can. Decay doesn't repeat on a calendar.

Everything below assumes you've cleared both of those, and that you're looking at one page that has been sliding for a month or more.

Recoverable by rewriting the page

These are the two causes where the classic advice — go update the post — is the correct advice.

1. Someone overtook you

The most common one, and the most boring. Your page didn't change; the set of pages it's judged against did. Somebody published something more complete, or an existing competitor improved theirs.

The test: search your own query in a private window and read the results above you. If the pages that now outrank you cover things yours doesn't — a section you skipped, a step you glossed, data you don't have — that's your answer, and it's visible in about four minutes. In Search Console the accompanying signature is clicks down, impressions down, average position at least a point worse.

The fix: an actual refresh of the substance, not the packaging. The content refresh checklist covers what to change and, more usefully, what to leave alone.

Cost of being wrong: low. This is the failure mode most people default to, and a genuine improvement to a page rarely hurts. It just may not be why you lost, in which case you spent a Saturday and the clicks don't come back.

2. The query changed shape

Subtler, and more often missed. The words people type are the same; what they want back has changed. A query that used to reward a 2,000-word explainer now returns a comparison table, a tool, a video, or six product pages. Your page is still good — it's answering a question the results page has stopped asking.

The test: look at the format of the top five results, not their content. If they are structurally a different kind of thing than your page, you have an intent shift. This is also the one where your average position can hold steady for a while before it goes, because Google is still testing you against the new set.

The fix: change what the page is, not how thorough it is. Adding another thousand words to a narrative guide when the results page wants a table makes the mismatch worse.

Cost of being wrong: moderate. Restructuring a page that didn't need restructuring can lose you the thing that was working.

Recoverable — but not by rewriting the page

Here's where the standard advice starts doing damage, because all three of these look like a content problem and none of them is one.

3. The results page started answering the question

You're winning the ranking and losing the click. An AI Overview, a featured snippet, a People Also Ask block or a screen of ads is satisfying the query before anyone reaches your result. Ahrefs analysed 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.

The test: the cleanest signal there is. Average position held or improved while clicks fell. Then search the query and look at the page — you'll see the reason occupying the top of the screen.

The fix: partial at best, and honesty matters more here than effort. Accept that some of that traffic is gone regardless of what you write, then decide whether the query is still worth serving. Usually the move is to push the page toward what a two-paragraph summary can't deliver: your own numbers, a real walkthrough, a decision the reader still has to make themselves.

Cost of being wrong: the highest on this list. Rewriting a page to win back clicks that a machine is intercepting is an expensive way to learn that lesson, and people do it repeatedly because the page "feels" like the problem.

4. You did it to yourself

Two of your own pages splitting the same query. A migration that changed URLs. A noindex that shipped with a template. A consolidation that redirected the wrong way round. An internal link you removed in a redesign without noticing what it was holding up.

The test: filter Search Console to the query the page is losing and open the Pages tab. If another of your URLs is showing up there, that's cannibalization — here's how to confirm it before you act on it, including Google's own position that there's no cannibalization penalty, only a picking problem. For the technical variants, run the page through the URL Inspection tool and see what Google actually has.

The fix: cheap, usually. Consolidate or differentiate the two pages, restore the link, remove the tag. Almost never a rewrite.

Cost of being wrong: low to check, high to skip. This is the one cause that's entirely within your control and it's the one people check last, after the weekend of writing.

5. Your links died

A site that linked to you went offline, or redesigned, or quietly dropped the placement. If the page's ranking was propped up by a handful of referring domains, losing them moves you.

The test: and here I'd rather be straight with you — this is the one cause you can't properly test for free. Search Console's Links report shows what it knows about now, and Google's own documentation says the report "includes links that Google has found over time" and that those links "may have since been removed," on top of being a sample rather than a full list. There's no compare mode and no lost-links view. So treat this one as a suspicion you can raise, not a finding you can confirm, unless you're paying for a backlink tool.

The fix: replacing links is outreach work. It's real, it's slow, and it isn't writing.

Cost of being wrong: moderate, and mostly wasted time — you can spend weeks chasing links for a page that lost for a different reason entirely.

Not recoverable

6. Demand left

Fewer people are searching. You are ranking exactly as well as you ever did for a thing the world has stopped asking about. This is the pattern most likely to get a perfectly good page needlessly rewritten.

The test: impressions down, average position held or improved, clicks falling roughly in proportion to impressions. Confirm it in Google Trends — Google's own traffic-drop guidance points you there for exactly this, to check whether a decline is yours or the whole topic's.

The fix: there isn't one. Refreshing a page cannot create searches. Either fold it into a broader page that still earns demand, or leave it alone and spend the afternoon on a page from cause 1, where the traffic is genuinely recoverable — when not to refresh a post is the whole decision, including the two moves that aren't refreshing and aren't nothing.

Cost of being wrong: you burn your best content hours on the one page on your site where effort has no upside. Which is why I treat this as the verdict whenever it shows up alongside anything else.

Why it's your best posts

The title of this piece isn't an accident, and the mechanism is worth naming, because it isn't bad luck.

Your best posts rank for your most competitive queries — which are the queries competitors are actively targeting, and the ones most likely to attract an AI Overview, because head-term informational queries are exactly what those are generated for. They also have the furthest to fall: a page earning 500 clicks a month can lose 200 of them, and a page earning 15 cannot. Clicks at stake scales with what you had.

And then the quiet one: your best posts are the pages you're least likely to touch, precisely because they're working. Nobody opens a well-performing post to check whether the results page changed underneath it. So the decline runs for months before anything makes you look — which is the actual reason decay is expensive. Not the rate. The delay.

Which one do you check first?

Not the likeliest — the cheapest to rule out. In this order:

1. Look at the shape and the breadth (30 seconds). One page or many? Slope or cliff? A cliff sends you to the technical causes, not this list.
2. Look at average position (1 minute). Position worse → cause 1 or 2. Position held or improved → cause 3, 5 or 6. That single number cuts the list in half.
3. Search the query (4 minutes). An AI Overview or snippet sitting on it → cause 3. A different kind of result winning → cause 2. Better versions of your page → cause 1.
4. Check impressions (1 minute). Down while your position improved → cause 6. Check Trends before you accept it.
5. Check yourself (5 minutes). Filter to the query, look for a second URL of your own. Inspect the page.
6. Links last. It's the least testable and the slowest to act on.

Six minutes of that beats an afternoon of rewriting, roughly every time. If you want the click-by-click version of reading those numbers, what content decay looks like in Search Console walks the five signatures and the thresholds behind them.

You've named the cause. Now what?

Two steps sit behind the diagnosis. Rank the pages before you fix any of them — run the full sweep (how to find declining pages in Search Console) and start with the page bleeding the most clicks per month, not the biggest percentage drop: (baseline weekly clicks − recent weekly clicks) × 4.33. That's which blog posts to update first.

Then prove it worked — compare the 28 days after against the 28 days before. If you guessed the cause wrong, that measurement is the only thing that will ever tell you.

Find out which pages are losing, and why

Everything above runs in free Search Console and I'd rather you knew how to do it than rented something you can't explain. The hard part isn't the six minutes — it's noticing in month two instead of month nine, on a page you had no reason to open.

That's the job RefreshRadar does: it watches every page in your Search Console, applies the same rules (20% below an eight-week baseline, above the 10-clicks-a-week floor), and emails you the ranked list — each page tagged with what its numbers did, in plain English, and the clicks per month at stake. It won't tell you a competitor outwrote you or that a link died; those are the four minutes of looking described above. What it does is put the right page in front of you, with the query it's losing, so you're not guessing which of the six you're even looking at.

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 get your three worst-decaying pages with the reason on each. If nothing on your site is decaying, we'll tell you that instead.