Most content management attention goes to getting something published. The lifecycle is everything that happens before and, more importantly, everything that happens after — and it is where large content libraries quietly fail. Publishing is a solved problem. Knowing that a document written four years ago is still accurate is not.
This article follows content through time. Which individual tasks can run unattended is covered in AI content automation; how people and AI divide the editorial work is in AI-assisted editorial workflows.
The stages, and where things actually break
A content lifecycle has eleven recognisable stages. Only a few of them are usually managed.
- Planning — deciding what needs to exist.
- Creation — producing it.
- Review — checking it before publication.
- Approval — a named person accepting responsibility for it.
- Publication — making it available.
- Distribution — delivering it to the channels that need it.
- Measurement — observing whether it works.
- Updating — keeping it accurate.
- Expiry — marking it no longer current.
- Archival — retaining it without presenting it as current.
- Deletion — removing it when retention allows.
Stages 1–6 receive nearly all the process attention because they are the path to getting something live. Stages 7–11 are where the value leaks. A library where nothing is ever expired accumulates content that contradicts itself, and the practical consequence is that people stop trusting any of it — including the parts that are correct.
What has to be true before AI helps
Lifecycle management depends on three pieces of information existing per item. No amount of AI substitutes for them:
- An owner. A named person, not a department. Unowned content cannot be reviewed, because nobody can decide whether it is still true.
- A review interval. How long this kind of content stays trustworthy. A pricing page and a company-history page are not on the same clock.
- A retention rule. How long this must be kept, which is frequently a legal or contractual answer rather than an editorial one.
These are usually derived from the content type, which is why classification matters here — knowing something is a policy rather than a blog post is what lets a retention rule apply at all. See AI content classification.
Where AI triggers fit
The pattern throughout is the same: AI detects a condition and raises it; a person decides what to do.
Review is due
The simplest and most valuable trigger. Content past its review interval is surfaced to its owner. No intelligence required — but pairing it with a summary of what changed nearby makes the prompt actionable rather than another ignored notification.
The content has probably gone stale
More useful than a calendar. Signals include: a referenced policy or price changed elsewhere; a newer page now contradicts this one; the product or service described has been withdrawn; search traffic to the page collapsed. Any of these suggests review is needed sooner than the interval says.
Duplication has appeared
Semantic similarity can find pages that now say substantially the same thing — usually because two teams solved the same problem independently. Left alone, these split search relevance and drift apart until they disagree.
Nobody uses this
Content with no traffic, no internal links, and no search hits over a long window is a candidate for retirement. Note that "unused" is not the same as "wrong" — a regulatory notice may be rarely read and must still be there. The trigger prompts a decision; it does not make one.
Expiry is approaching
Where content has a known end date — a campaign, a contract term, a dated notice — the platform should surface it before it becomes wrong, not after someone complains.
Expiry, archival and deletion are three different things
Conflating them causes real problems, and the distinction is worth being pedantic about:
- Expiry means no longer presented as current. The content may still be reachable, but it should stop appearing in search and navigation as though it were live, and it should not be retrievable as a source for an assistant's answer.
- Archival means retained deliberately, out of the active library. Findable by someone who needs the historical record, invisible to someone looking for current guidance.
- Deletion means gone. This is a records decision governed by retention obligations, and it is the one stage that should never be fully automated.
The middle state is the one most platforms lack, and its absence is why organizations either keep everything — and lose trust — or delete things they later need.
A URL is a commitment
Retiring content is also a URL decision. A page that has existed for years is linked from elsewhere, bookmarked, and indexed. Removing it without a redirect turns every one of those into a dead end. Expiry and archival should preserve or redirect the address even when the content moves out of the active library.
Sitting on a library nobody has reviewed in years? Schedule an AI CMS consultation — the inventory is usually more revealing than people expect.
Measurement is part of the lifecycle, not a separate exercise
Stage 7 is skipped more than any other, and it is what makes stages 8–11 decidable. Useful signals are mostly already available: page traffic, internal search queries that led here, searches that returned this and were abandoned, support contacts on this topic, and the proportion of a content set currently within its review period.
That last one is worth tracking as a standing measure. It is a single number that describes whether the library is being maintained, and it degrades slowly enough that nobody notices without it.
Governance controls that belong here
Lifecycle work touches records obligations, so a few controls are not optional: versioning so a prior state can be recovered, an audit trail of who changed status and when, retention rules that reflect actual obligations rather than habit, and evidence of review — a record that someone confirmed the content was still accurate on a date.
That last point is what distinguishes a maintained library from one that merely looks maintained. These sit inside the wider framework described in AI content governance, which is where accountability for the cycle is assigned.
How LABUSA approaches it
We start with an inventory: what exists, who owns it, when it was last reviewed, and whether anyone will vouch for it. The answer is frequently uncomfortable, and it usually reorders the project — a large share of what organizations plan to migrate turns out to be content that should be archived instead.
From there we implement the lifecycle in the platform rather than in a spreadsheet: ownership and review intervals as fields, triggers that surface work to the right person, an archival state distinct from deletion, and retention rules recorded against content types. Our enterprise content platform work treats this as core scope, because a platform without it degrades from the day it launches.
Frequently asked questions
How often should content be reviewed?
By content type rather than uniformly. Regulatory and pricing content on a short cycle, service descriptions in the middle, historical and archival material rarely. A single interval across everything either exhausts the team or misses what matters.
Can AI decide what to delete?
No. It can identify candidates and evidence — unused, superseded, past retention — but deletion is a records decision with legal weight, and it needs a person.
What if nobody owns most of our content?
Then assigning ownership is the first deliverable. It is an organizational conversation rather than a technical one, and no lifecycle process functions without it.
Should expired content be removed from search?
From the current-content search, yes — and from an assistant's retrievable sources, definitely. Keeping it reachable through an explicit archive view is usually the right balance.
Where does this fit against a migration?
Before it. Deciding what to archive is the cheapest part of any migration, because it reduces everything that follows — see how to implement an AI-powered CMS.
Related reading
- AI Content Automation — which lifecycle tasks can run unattended.
- AI-Assisted Editorial Workflows — the people side of review and approval.
- AI Content Classification — content type drives review intervals and retention.
- How to Implement an AI-Powered CMS — where lifecycle design fits in the project.