Why content decays
Decay is not a sign of bad content; it is what happens to all content. A guide that was accurate at publication describes a product that has since changed, cites a rule that has been replaced, links to pages that no longer exist and uses screenshots of an interface that has been redesigned. Meanwhile new pages are published on nearby topics, some of them contradicting the old ones, and the library becomes a place where a reader can find two answers to the same question and no way to tell which is current. Search engines notice too: Google's guidance on helpful content asks whether a page is accurate, whether it shows real expertise and whether the site as a whole is trustworthy, and a library full of stale and duplicated pages undermines the E-E-A-TExperience, Expertise, Authoritativeness, Trust. signals the E-E-A-T guide explains.
The cost is spread thin and so goes unnoticed. Each stale page loses a little trust, a little traffic and a little conversion; support receives questions the content used to answer; writers spend time rewriting pages that already exist. None of it appears in a report as a single line, which is why governance has to be a standing routine rather than a response to a problem. The content strategy hub places governance as the fourth strategy decision for that reason: it is the one that protects the value of the other three over time.
Standards: style, accuracy, accessibility
Standards are the rules every page is held to regardless of who wrote it. Three kinds matter, and each should be short enough that people actually read it. They sit inside the team's wider editorial standardsThe rules for how content is made and kept honest: who writes and reviews it, how claims are sourced, how it is funded, where AI helps and how it is corrected., which also say how content is funded and how errors are corrected.
Style standards cover voice, spelling conventions, formatting, how headings and links are written, and the house positions on contested points (how the product is named, which terms are avoided, how competitors are referred to). A style guide of a few pages, with examples, does more than a long one nobody opens. The tutorial on writing an editorial style guide builds one from the corrections editors already keep making, with a right and a wrong example under every rule.
Accuracy standards say what a claim needs before it is published: a source for any figure or external fact, a named reviewer for anything technical, legal or financial, a date on anything that will age, and a rule that figures appear only where they can be attributed. They also say what happens when a fact changes — who is told, which pages are checked, how fast. For search-facing pages the accuracy standard overlaps with SEOImproving visibility in unpaid search results. practice, since experience and verifiable expertise are part of what search quality guidance looks for.
Accessibility standards make content usable by people with disabilities and by assistive technology: headings in order, descriptive link text rather than "click here", alternative text for informative images, captions and transcripts for audio and video, sufficient contrast, and plain language. These are writing standards as much as design standards, which is why they belong in governance rather than in a technical checklist alone.
Roles and approvals
Most review problems are role problems: nobody knows who decides, so drafts wait, or everybody decides, so drafts bounce. A governance model names the roles and the stage each one acts at. Small teams combine roles in one person; the point is that each role is assigned, not that each has a separate person.
| Role | Responsible for | Acts at | Typical holder |
|---|---|---|---|
| Content owner | The page's accuracy and currency over its life; the review date | Ongoing | Writer, product marketer or subject lead |
| Editor | Style, structure, clarity and fit with the brief | Draft review | Content lead or managing editor |
| Subject reviewer | Technical, legal or financial accuracy of claims | Before approval, on flagged pages only | Named expert per pillar |
| Approver | Final sign-off for pages with regulatory, legal or brand weight | Before publish, on defined page types | Head of marketing, legal, compliance |
| Publisher | Formatting, metadata, links, accessibility checks, scheduling | Publish | Content operations or web editor |
| Governance owner | The inventory, the refresh calendar, the standards and the audit | Continuous | Content lead |
Approvals should be proportionate. A short social post from an established series needs an editor; a pricing page or a claim about a regulated product needs a subject reviewer and an approver. Writing down which page types need which approvals removes the most common delay, which is a draft waiting for a sign-off nobody realised was optional. The same discipline applied to reporting assets — one owner per dashboard, defined reviewers, a retirement rule — is described in the dashboard governance guide, and the two routines can share a calendar.
The content inventory and audit
The inventory is the list of everything published, with the fields governance needs: URL, title, pillar, stage, format, owner, publish date, last review date, next review date, status and the decision from the last audit. It is the single artefact that makes the rest of governance possible, because a review cycle cannot run over pages nobody has listed. Build it from the site map and the content management system, enrich it with performance data, and keep it where the team already works rather than in a file that is opened once a year. The content inventory and audit template lays these fields out as columns, with the evidence, the decision and the destination for retired pages beside each one. When the business enters a new market, the same inventory is the starting point of the tutorial on deciding which content to localise first, which sorts it into a first wave, later waves and content left out.
An audit reads the inventory against evidence and ends in a decision per page. The evidence comes from two places: performance, grouped by pillar and stage as described in measuring content performance, and a human read for accuracy, duplication and fit with the current strategy. Each page gets one of four decisions — keep, refresh, consolidate or retire — with a reason and a date. The audit is also where the library's KPIA metric tied directly to an objective. is read at the level that matters: the share of pages reviewed on schedule, the share with a current owner, and the number of pages that contradict another page.
A quarterly audit of a software company's help and blog content
The content lead exports an inventory of 640 pages and joins it to the last 12 months of analytics. 310 pages earned fewer than 20 sessions in the year; 85 of those sit at the decision stage and are kept for sales use, the remaining 225 are flagged for consolidation or retirement. A read of the 60 highest-traffic pages finds 14 with out-of-date screenshots and 6 citing a pricing tier that no longer exists. The audit ends with 410 keep, 96 refresh, 88 consolidate into 31 surviving pages, and 46 retire with redirects to the nearest replacement. The refresh work is scheduled into the next two quarters at 8 pages a month, and 38 pages without an owner are assigned one before anything else happens.
Refresh, consolidate, retire
Refresh is the default for a page that still answers a question people ask. It means re-reading the page against the current product, rules and sources; updating facts, links, screenshots and dates; checking that the format still matches what the question wants; and recording the review in the inventory. A refreshed page keeps its URL and its history, which is why refresh is usually preferable to a rewrite under a new address.
Consolidate applies when several pages answer the same question, or when thin pagesPages that offer little original value of their own, such as near-empty templates, near-duplicates or copies with only a word or place changed. together would make one useful one. Choose the page with the strongest history and best fit, merge the useful material from the others into it, and redirect the retired addresses to it so that readers and links arrive at the survivor. Consolidation is where most of the "two answers to one question" problems are solved. When a reader or a colleague points out two pages giving different answers between audits, the diagnosis of conflicting information across your pages settles which version is true and traces the route that let the pages split.
Retire applies when a page answers a question nobody asks any more, describes something that no longer exists, or cannot be made accurate at reasonable cost. If anything still points at it — links, traffic, a close replacement — redirect it; if nothing does, remove it and let it return a not-found response. Either way the inventory records the decision and the date, so that the next audit does not rediscover it. Retirement is the decision teams avoid most and the one that improves a library fastest.
The pages a platform generates around the library, such as tag, author and date archives, need the same decisions made once per template, which the index rules for tags, archives and utility pages set out.
Governance for AI-assisted content
AI-assisted drafting changes the volume and speed of production without changing what governance has to guarantee: that every published claim is accurate, sourced where it needs to be, original, in the brand's voice and reviewed by a person who takes responsibility for it. Governance therefore adds rules rather than a separate process. Which stages may use AI assistance and which may not. What the reviewer must check before a draft that used AI reaches the editor: facts against sources, quotations against originals, originality, and any invented reference. Who is accountable for the page — always a named person, never the tool. How the use is disclosed, where the organisation has decided to disclose it. And how AI-assisted pages are flagged in the inventory so that audits can check whether they age differently.
The AI-assisted content with human review guide sets out the production workflow and the review gate in detail; governance owns the policy those gates enforce, the record of which pages went through them, and the periodic check that the gates are being used rather than skipped under deadline. The policy itself can be short, and the step-by-step guide to writing a marketing AI use policy shows how to draft it from a register of the uses the team already has.
When drafts queue for sign-off and the calendar slips, the page on content production that is too slow shows how to measure each wait and which role or stage to fix first.
For a library that is already sliding, the diagnosis of old content losing traffic gradually sorts each decaying page into stale facts, overlapping posts, fading demand, stronger competitors or lost internal links before anything is refreshed.
A governance process that funds a refresh cycle can check that the cycle earns its keep with the content refresh test design.
Common mistakes
- Governance as a project. An annual clean-up without an ongoing routine; the library decays between projects.
- No owner per page. Pages owned by "the team" are reviewed by nobody.
- Standards nobody reads. A long style guide is ignored; a short one with examples is used.
- Approvals for everything. Uniform sign-off delays the routine work and still misses the pages that need scrutiny.
- Deleting instead of redirecting. Retired pages with links and traffic strand readers and lose value.
- Auditing without decisions. An inventory with performance columns and no keep, refresh, consolidate or retire verdict changes nothing.
- AI as an exception. Treating AI-assisted pages as outside the standards, or as needing no named accountable person.
- Accessibility left to design. Heading order, link text and alternative text are writing decisions; a page can pass a technical check and still be unusable.
