The Blog That Stopped Paying: Auditing, Refreshing and Pruning Content with Semalt

There is a particular moment in the life of every content programme. Two or three years in, the blog has ninety articles, traffic peaked eighteen months ago, and nobody can explain what happened because nothing obviously broke. Nothing did break. The content simply aged, the results pages changed around it, and the decline was slow enough that no single month looked alarming.

This article is about handling that systematically with the Semalt platform: how to find decaying pages, how to decide what to do with each one, and — the part almost everybody skips — how to tell afterwards whether the decision was right.

Why content decays even when nothing changed

Four separate mechanisms produce the same symptom, and the correct response differs for each.

Distinguishing the third and fourth rows from the first two is the entire skill. Teams that skip this step spend months rewriting excellent articles about topics nobody searches for any more, and conclude that content refresh does not work.

Working parameters from the process described below.

Finding the pages that matter

The instinct is to sort by biggest decline. That produces a list dominated by pages that never mattered, because a page going from forty visits to four shows a ninety per cent drop and represents nothing.

Sort instead by lost value. For each URL, you want three things visible together: what it earned at its peak, what it earns now, and whether the queries behind it still have demand. The pages worth your quarter are the ones that were genuinely valuable and are now visibly less so, while the topic is still alive.

1. Pull position and impression history per URL

   Twelve to eighteen months, so you can see shape rather than a snapshot. A page that dropped and recovered is a different case from one on a steady slide.

2. Separate query decline from page decline

   If impressions fell for everyone ranking on the topic, demand moved. If they fell only for you, you lost ground to someone. Only the second case is a content problem.

3. Check what now ranks above you

   Open the results page. If the top three are a different content type than yours — a comparison, a tool, a video — no amount of rewriting your format will fix it.

4. Tag each page with one of four decisions

   And write the reason in one line. This is the step that makes the audit finishable, because it converts a list of problems into a queue of tasks.

When to update, and what "update" means

An update that works is not a new publication date and a tidied paragraph. Search engines are unimpressed by cosmetic freshness, and readers notice it immediately.

A real update replaces what has actually gone out of date: figures, screenshots, tool names, regulatory details, examples that reference a version of the world that no longer exists. It also usually means restructuring, because the article was written to answer a question that has since been reframed by the results page around it.

Keep the URL. This is not negotiable — a refreshed article at a new address discards every link and every accumulated signal the original had, which is the most common way a well-intentioned refresh makes things worse.

Merging: the underused decision

Most ageing blogs contain three or four articles that answer roughly the same question from slightly different angles, written months apart by people who did not know the others existed. They compete with each other, none of them is complete, and all of them are mediocre.

We cover this in detail in B2B SEO with Long Sales Cycles.

Merging is usually the highest-return action available. Take the strongest URL, absorb the useful material from the others, redirect the weaker URLs to it, and update the internal links. One thorough page consistently outperforms four partial ones, and the work is mostly editing rather than writing.

Deleting: when nothing is worth saving

Some pages should simply go. Announcements from four years ago, event pages for events that happened, thin posts written to hit a monthly quota, duplicated category descriptions. They add nothing for readers and they dilute the crawl.

There is a full walkthrough in Four Hours a Week.

The last item on the right catches people out regularly. A page about a service that only sells in autumn looks dead in April, and deleting it in April is how you discover in September that you have to start again.

Measuring whether the decision worked

Refresh programmes fail most often not because the edits were wrong but because nobody checked. Six months later there is no way to tell whether the work helped, so the next cycle is decided by instinct again.

The measurement is straightforward if you set it up before starting. Record the position and impression baseline for each refreshed URL on the day you publish the update, then check at four and eight weeks. Four weeks tells you whether anything moved; eight tells you whether it held.

Two things make this reliable. Refresh in batches with a clear boundary — ten pages in one week, then nothing for a month — so you can attribute movement rather than guessing across a continuous trickle of edits. And keep a simple log of what was changed on each URL, because "we updated it" tells you nothing eight weeks later, while "moved the answer up, added the price table, merged in two old posts" tells you exactly what to repeat.

A sustainable cadence

The failure mode of content audits is ambition. A team decides to review ninety articles, does eleven, loses momentum, and the spreadsheet is never opened again.

A quarterly batch of ten pages is boring and finishes. Over two years that is eighty pages reviewed, which for most sites is the entire meaningful archive, done without ever declaring a special project. Between batches, the only ongoing task is watching for new decline — which the platform surfaces as position movement, so nobody has to remember to look.

See also: Redesigning Without Losing Google.

One more habit worth building: when publishing anything new, note the two or three existing pages it should link to and from. Most decay is accelerated by articles being orphaned as the site grows around them, and five minutes of internal linking at publication prevents years of slow decline afterwards.

A note on smaller language markets

Everything above applies more sharply in Portuguese than in English, for one structural reason: the archive is smaller and the competition thinner, which cuts both ways.

The favourable side is that a genuinely thorough page in Portuguese often faces four or five competitors rather than four hundred. Refreshing a decayed article to be clearly the best available answer is realistic here in a way it frequently is not in English, where the same effort buys a place on page two. For a Portuguese-language site, refresh usually has a better return than new publication, and most teams have the ratio backwards.

The unfavourable side is measurement. Lower volumes mean noisier data: a page that goes from twelve impressions a day to seven has not necessarily declined, it has fluctuated. Judge Portuguese-language decay over longer windows — quarter against quarter rather than month against month — and be sceptical of any conclusion drawn from a page with very low absolute numbers. The pattern you are looking for is a sustained slope over months, not a dip.

There is also a translation trap specific to bilingual sites. When the Portuguese article decays and the English one does not, the instinct is to translate the English version over the Portuguese. That almost always makes it worse, because the two versions serve different queries with different competitors. Refresh each language against its own results page, or you end up with a fluent Portuguese article optimised for an English-language search behaviour that does not exist here.

The honest summary

Content refresh is not a growth tactic; it is maintenance. It rarely produces the dramatic charts that new content can, and it is consistently cheaper per unit of traffic recovered than writing something new, because the page already has history, links and a proven topic.

What makes it work is the discipline of assigning one of four decisions to every page and then actually measuring the result. What makes it fail is treating every declining page as something to rewrite, which guarantees the audit never ends.

If your archive has never been reviewed, start with the ten pages that were most valuable at their peak. Open the dashboard, pull the position history for those URLs, and check whether the topic declined or you did. That distinction decides everything that follows.