Bloom is now Rainbrand

By Siddharth Borderwala

Published

12 min read

What is content governance? A framework, a plan template and the AI version

Content governance decides who creates, approves, publishes, maintains and retires content, and by what standards. A six-part framework, the NHS website's approval chain as a worked example, a plan template to copy, and what changes when AI agents draft and publish.

TL;DR

Content governance is the set of decisions about who may create, approve, publish, maintain and retire content, which standards it must meet, and how you check that it does. A workable framework names roles, standards, a workflow matched to risk, a lifecycle with review dates, regular audits and a few metrics. AI agents are a new kind of author inside that framework, not a reason to replace it.

What content governance is

Every organisation that publishes ends up with content nobody owns: a pricing page with last year's prices, three help articles that disagree, a blog post about a retired product. Content governance is the set of decisions that prevents that. Contentful defines it as “the big-picture management of content across an organization”: the processes, policies and standards that keep content on-message, compliant and good enough to publish.

In practice it answers five questions for every kind of content you publish. Who can create it? Who has to approve it? What standards must it meet? How long is it valid, and who checks it after that? How do you know the whole system is working? Frances Carden of the Library of Congress put the stakes plainly on Digital.gov: “Unmanaged content is a liability.”

It sits next to two related ideas. Content strategy decides what to publish and why; governance decides how it gets made, checked and kept accurate. Brand governance is narrower: who sets and changes the brand's own rules, and how agents are kept within them. This article covers all of an organisation's content, with the brand, and the brand system that records it, as one of its standards; brand management covers how that brand is kept current.

The content governance framework

Simpplr's glossary lists ten elements; most frameworks reduce to six parts. Each part produces something you can point to:

PartWhat it decidesWhat it produces
RolesWho requests, creates, reviews, approves, publishes and owns each content typeA role table, with a named owner per content type
StandardsWhat content must meet: brand and voice, accuracy, legal, accessibility, SEOStyle guide, brand guidelines, claims list, checklists
WorkflowThe steps from request to publication, and how many checks each type needsA workflow per content type, scaled to risk
LifecycleHow long content stays valid, when it is reviewed, when it is retiredReview dates and retirement rules
AuditsWhat is actually published, and whether it still meets the standardsAn inventory and a regular audit
MetricsWhether governance is workingA short dashboard: errors, review times, stale pages

The lifecycle is the part that is easiest to underbuild. Publishing is easy to govern because someone is watching. Six months later, nobody is, and that is when product names change, claims go out of date and pages contradict each other.

Content governance models

The first decision is where authority sits. There are three common answers, and each works for a different kind of organisation:

ModelHow it worksA real example
CentralOne team writes or commissions the content and signs it off before it is publishedThe NHS website, where a dedicated content team produces the pages and an internal clinical team approves every clinical one
HybridA central team owns standards, templates and key pages; units publish within themThe University of Washington, whose web strategy team is the central authority while units run their own sites
DecentralisedMany contributors publish directly, under shared policies they help writeWikipedia, whose policies are “developed by the community to describe best practices”

Each model trades something. Central sign-off sends every page through one approval process, at the cost of a queue that someone has to staff; the NHS accepts that because a wrong health page can hurt someone. Decentralised publishing scales, but depends on shared standards and people who enforce them. Many organisations with several publishing teams settle on a hybrid: a small central team owns the standards, the templates and the highest-risk pages, and everyone else publishes inside those lines.

AI agents fit any of the three, but they change the arithmetic. If agents let every team publish much more, a central approval queue grows with them, which pushes organisations towards hybrid models with stronger automated checks.

A central approval chain, written down

The NHS website publishes its content policy, and it is unusually specific about who does what. Here is how its policy maps to a governance plan for one content type, a new page about a health condition:

Part of the planWhat the NHS website's policy says
CreatorA dedicated content team, trained to find and review high-quality evidence
Editorial checkA senior, experienced member of staff checks new content and major changes for accuracy, balance, accessibility and tone
ApproversTwo members of the internal clinical team for new content; a policy official at the Department of Health and Social Care or NHS England when a policy applies
EscalationContent about public health emergencies or fast-changing policy goes to NHS Digital's Chief Medical Officer, or someone they delegate to. That is the policy's wording: NHS Digital merged with NHS England on February 1, 2023, after the page was last reviewed
Accountable ownersThe programme's Clinical Lead for clinical safety and evidence; the Content Director for editorial standards
EvidenceSources such as NICE and Clinical Knowledge Summaries, available on request; publication dates shown on every page
Review triggersAt least every three years, plus evidence updates and user feedback, considered as they arrive
Retirement actionPages are amended as soon as evidence or feedback requires it; on its YouTube channel, videos that no longer reflect current evidence are suspended, then updated, replaced or removed

Two things stand out. Accountability is assigned to named roles, separately from who writes the page. And the lifecycle is part of the policy, not an afterthought. It is also hard to keep: when we read the policy in November 2026, its own footer said the next review was due on October 14, 2025.

Who does what

Whatever the model, each step needs a named role. In a small team one person may hold several; the point is that nobody has to guess who decides.

StepUsually done byWhere agents fit
RequestThe team that needs the contentAn agent can open a request, for example when a product changes
CreateWriters, designers, subject expertsAgents draft for a named person, who answers for the result
ReviewAn editor, plus a subject expert for factsAutomated checks for style, brand and links before a person reads
ApproveThe content type's owner; legal for regulated claimsA person, for anything medium or high risk
PublishWhoever holds publishing rights in the CMSAgents can publish low-risk content where the plan allows it
MaintainThe owner, on the review dateAgents can flag pages past their date or out of line with a change
RetireThe owner, with the web team for redirectsA person decides; agents can prepare the list

The “Where agents fit” column is a starting point, not a rule, and running content operations with AI covers each stage in more detail. Write your own version into the plan, by content type.

Lifecycle and audit cadence

Every content type needs a review interval, the events that trigger an early review, and a rule for when it is retired. These are reasonable starting points; adjust them to how fast your products and regulations change.

Content typeReviewRetire when
Pricing and plansEvery quarter, and on any price or plan changeThe plan no longer exists
Product and help pagesEach release, and when a feature, name or screen changesThe feature is removed
Legal and regulated pagesEvery year, and when a law or policy changesA new version replaces it
Blog posts and guidesEvery year, and when a product, claim or cited source changesNot worth updating; redirect it
Campaign pages and adsWhen an offer or claim changes mid-flightThe campaign ends
Sales decks and templatesEvery six months, and on positioning, pricing or logo changesA new template replaces it

Audits check what is actually out there against these rules. A full inventory once a year is enough for most sites; a monthly report of pages past their review date does more day to day. Digital.gov's advice is blunt: use analytics to find pages with little or no traffic, and if a page gets almost no views, “it might be ready to retire”.

Metrics that show it is working

Governance that can't show results gets bypassed. Track a handful of numbers:

MetricWhat it tells you
Errors found after publishing, per monthWhether review is catching problems before readers do
Share of content past its review dateWhether the lifecycle is actually running
Median time from draft to approvalWhether governance is slowing teams down
Share of content with a named ownerWhether roles exist outside the plan
Repeated findings in reviewsWhich standard is unclear, or which source is missing
Content retired or merged per quarterWhether the inventory is shrinking as well as growing

If review times climb, people will find ways around the process. Watch the third row as closely as the first.

A content governance plan you can copy

Fill this in for your organisation. Start with the content types that carry the most risk, such as pricing, product claims, regulated statements and help content, and add others later.

CONTENT GOVERNANCE PLAN

1. Scope
   Content types covered:
   Out of scope:
   Governance model (central, hybrid,
   decentralised):

2. Roles, per content type
   Requester:
   Creator (person or agent):
   Reviewer:
   Approver:
   Owner after publishing:

3. Standards (link each, name its owner)
   Brand guidelines and voice:
   Style guide:
   Approved claims, with sources:
   Legal and regulated statements:
   Accessibility:
   Search and metadata:

4. Workflow by risk
   Low risk: checks before publishing:
   Medium risk: checks and approver:
   High risk: checks, approver,
   sign-off record:

5. Lifecycle, per content type
   Review interval:
   Early-review triggers:
   Retirement rule and who decides:

6. Audits
   Inventory location:
   Cadence and owner:
   What an audit checks:

7. Metrics
   Errors found after publishing:
   Content past review date:
   Median draft-to-approval time:

8. AI-made content
   Accountable person for each agent:
   Types agents may draft:
   Types they may publish alone:
   Where they get standards and facts:
   How AI use is recorded and disclosed:

Keep the plan short enough that people read it. A one-page role table and a risk-based workflow will do more than a forty-page policy nobody opens.

AI content governance

AI doesn't change what content governance is for. It changes the volume, and it adds a new kind of author that is fast, tireless and has no memory of last week's correction. Three parts of the framework need explicit answers.

Roles. An agent can draft, and publish where the plan allows it, but it can't be accountable. Each agent acts for a principal: a named person or team who sets its permissions and answers for what it publishes, the way the NHS's Clinical Lead answers for clinical safety whoever wrote the page. The plan above has a line for that.

Standards and facts. A person learns which claims are approved over months. An agent knows only what its task gives it, so standards and approved facts have to reach it with each task, and anything it adds beyond them should be treated as unverified. Newsrooms were early to write this down. The Associated Press's standards around generative AI, published in August 2023, say: “Any output from a generative AI tool should be treated as unvetted source material.” The Guardian's principles from June 2023 set an approval rule: significant AI-generated elements appear “only … with clear evidence of a specific benefit, human oversight, and the explicit permission of a senior editor.”

Workflow. Review depth should follow risk, not authorship. A low-risk social reply drafted by an agent may need only an automated check; a pricing page needs the same sign-off whether a person or an agent wrote it.

Here is what the facts rule looks like when it reaches an agent's task. Extropic is a chip company working on thermodynamic computing. In a published Rainbrand example, Claude made a two-page concept datasheet for Extropic's Z1 chip, using only the figures in Extropic's summer 2026 update. Rainbrand keeps a brand's guidance and files and returns what a task needs as a Brief. This one noted that the brand's files didn't contain the update, so the figures counted as task-supplied, and it set the rule:

For the requested “only figures in the update” constraint, do not add numbers, benchmarks, comparison claims, power conditions, or other details beyond the update/task-provided material; preserve qualifiers such as “over” and “under.”

The datasheet kept to it. The figures keep their qualifiers, the efficiency point is attributed (“Extropic says”), and the footer names the source:

Part of the Extropic Z1 concept datasheet: a specifications table with values such as over 50 MHz and under 1 W, a paragraph beginning Extropic says, and a footer naming Extropic's summer 2026 update as the source

The Z1 concept datasheet Claude made, with its source named in the footer

Rainbrand's Review, which checks an image of an asset against the brand, passed the datasheet's copy in both rounds, noting that the text matched the requested specifications. That compares the page with what the request supplied. Whether those figures match Extropic's update is a different question, and answering it means a person reading the source. The facts rule narrows what the agent may write; it doesn't replace that check.

A governance plan for what agents make comes down to a few rules:

  1. Give agents the standards and approved facts with each task. Fetch them from one maintained source rather than pasting a policy into each tool, where it goes stale.
  2. Treat anything the agent adds as unverified. New numbers, comparisons, quotes and claims go to the content's owner before publishing.
  3. Check every asset before it ships, at the depth its risk needs. Automated checks catch format, brand and copy errors and return findings to revise from; a person still decides on claims.
  4. Record what the agent was given and what it produced. When a page turns out wrong, you need to know whether the standard, the facts or the agent was at fault.
  5. Put AI-made content on the same review dates as everything else. It goes out of date just as fast.

FAQ

What is the difference between content governance and content management?

Content management is the software and day-to-day work of storing, editing and publishing content, usually in a content management system (CMS). Content governance is the set of decisions the CMS should enforce: who may publish what, which checks come first and when content is reviewed or retired. A CMS can hold the workflow, but it can't decide it.

How do you get teams to follow content governance?

Make the governed path the easy one. Put the standards inside the tools people already use, as templates, checklists and automated checks; set review dates in the CMS rather than in a spreadsheet; and keep approval queues short for low-risk content. When following the rules is slower than ignoring them, people ignore them.

What should a content audit check?

Accuracy against current products, prices and claims; broken links; outdated screenshots and names; brand and style; accessibility; and whether anyone still reads the page. The output is a list of pages to update, merge, redirect or retire, each with an owner and a date.

Do you have to disclose AI-generated content?

It depends on where you publish and what the content is. Some publishers commit to it voluntarily, as the Guardian does for significant AI-generated elements, and some laws and platforms require labels for certain content. Record AI use in your plan either way, and check the rules that apply to you.

All 9 Review articles
  • 11 min read

    AI governance for marketing: permissions, approvals and brand checks for agents

    What marketing agents may read, draft, change and publish in your CMS, ad accounts, email platform and CRM, which assets need whose approval, what to record and disclose, and where brand checks fit. With a proposed one-page policy for one workflow, and a run that shows a stale brand edit being refused.

  • 15 min read

    How to run brand compliance monitoring across partners and channels

    Brand compliance monitoring checks your brand after it ships, on your own channels and in the assets partners, resellers, franchisees and affiliates publish. What to sample and when, how to check what you capture, how to escalate and what to measure, with a monitoring log to copy.

  • 8 min read

    What is brand intelligence? The monitoring meaning and the production meaning

    Brand intelligence usually means data about how a brand is seen and discussed: listening, sentiment, share of voice, reviews and now AI answers. Vendors also use it for applying and checking brand knowledge while assets are made. What each meaning covers, the kinds of tools, and which one you need.

Connect Rainbrand.

  1. Open the Rainbrand plugin and add it.

    Open Rainbrand plugin (opens in a new tab)
  2. Connect your Rainbrand account when asked. It works in ChatGPT and in Codex.