Every post has a half-life. Most briefs never ask what it is.

A post that needs rewriting every quarter is not a failure. It is a cost you agreed to when you commissioned it, almost always without noticing. What decides how fast a post decays is not its subject but the kind of evidence you used to make it true, and that is knowable before a word is written.

What makes a claim true sets its clock

Take one subject and argue it three ways. You can argue from principle: what stalls a company blog is coordination, not writing time. You can argue from how a named system behaves right now: this is what a platform's documentation currently says. Or you can argue from a number: the share of buyers who do some particular thing.

The first is close to permanent. It rests on how organisations work, and organisations change on the scale of decades. The second decays the moment the vendor ships something, and vendors do not send notice. The third decays on the release schedule of whoever publishes the data, which at least has the courtesy of being a schedule.

Same subject, three different maintenance commitments, and the choice was made at the moment you picked your evidence. That choice is usually made for rhetorical reasons — a number feels more convincing than a principle — and the maintenance bill is never part of the conversation.

The question to ask before the post is written

The test fits in one sentence. What would have to happen in the world for this paragraph to be false, and does that thing have a schedule?

Some answers are comfortable. Nothing foreseeable, in which case the paragraph is durable and you can stop thinking about it. Or an annual statutory change, a quarterly filing, a yearly survey — a clock you can write down. A page with a known expiry date is not a problem. It is a calendar entry.

The uncomfortable answer is the third: it could change at any time and nobody will tell you. A competitor's pricing page. A platform's undocumented behaviour. An integration partner's API. These are the paragraphs that quietly turn false, and they are also, reliably, the most useful paragraphs in the post — the concrete, checkable things nobody else bothered to write down. We are not arguing you should avoid them. We are arguing you should know when you have written one.

Most of the cost is in finding out, not in fixing

Rewriting a section takes twenty minutes. Discovering that it needed rewriting takes a full read of the archive, and in practice it happens because somebody stumbled over the page — or because a customer did.

That is why an archive carrying no decay information is more expensive than its size suggests. Every audit starts from nothing, because you cannot sort by risk when nothing recorded which pages rested on a claim with a clock on it. So the audit gets deferred, and the pages that were always going to expire expire in public.

The fix is unglamorous. When the post is written, record what would falsify it and when that becomes plausible. Expiry then becomes a queue rather than an excavation, and the pages that need nothing can be left alone deliberately instead of neglected hopefully. We have argued that some subjects do not belong on a blog at all; this is the next question down, for the ones that do.

Changing the date is not maintaining the page

There is a cheap version of this, and Google names it directly. Its guidance on helpful, people-first content asks whether you are changing the date of pages to make them seem fresh when the content has not substantially changed. It asks the same about bulk-adding or removing content to make a site seem fresh, and answers that one itself: no, it won't help. The guidance is worth reading in full.

So a maintenance pass either touches the substance or leaves the date alone. That is the discipline, and it is the part automation handles worst. An engine can tell you a paragraph's clock has run out, because the brief recorded what the claim rested on. It cannot tell you what replaced the fact. That means going to a source, and a model that invents the new number is worse than a page that is honestly out of date — a stale page at least carries a visible date, and a reader knows how to discount one.

This is the cornerstone argument again. Google's objection is to content nobody needed, not to machine authorship, and a confident sentence that stopped being true is unhelpful no matter who typed it.

Budget the rewrite when you commission the post

None of this says avoid perishable claims. The perishable paragraph is often the reason the post is worth reading at all. The argument is only that it should be a decision rather than an accident.

So the brief gets one more line: what is the clock on this, and who goes and looks when it runs out. Most posts will answer none, and that is a real answer worth recording rather than a shrug. The handful that answer six months become a maintenance schedule instead of an ambush. And when you eventually reposition and find that your archive still keeps the old promise, you will already know which pages to read first.