CCMS and AI content, defined.

A Component Content Management System. Software that stores content as small, reusable components rather than whole documents, so the same approved piece can be used across many outputs and kept consistent everywhere.

Content written and stored as small, self-contained chunks (a warning, a step, a spec) that can be reused, updated, and assembled into many documents instead of being trapped inside one file.

Content that follows a defined model, where every piece has a clear type and place. Structure is what lets both people and machines find, reuse, and trust the right content.

One governed place where each piece of content lives. Update it once and every output that uses it updates too, so there is only ever one correct version.

Using the same approved component in more than one place instead of copying and pasting. High reuse means lower cost, faster updates, and consistency across every document.

A defined category of content with its own rules and structure, such as a procedure, a warning, or a reference topic. Consistent content types make reuse and automation possible.

The blueprint that defines the content types an organisation uses and how they fit together. A good content model is the foundation for reuse, governance, and AI-readiness.

How content is organised, labelled, and connected so people and systems can find what they need. Strong information architecture is the difference between a library and a pile.

A consistent set of labels and categories applied to content so it can be filtered, found, and retrieved reliably. Taxonomy is the metadata backbone of search and AI retrieval.

The stages content moves through: created, reviewed, approved, published, maintained, and eventually retired. Managing the lifecycle is how content stays accurate over time.

A CMS manages whole pages or documents, usually for a website. A CCMS manages content at the component level for reuse across many outputs and channels, which matters when accuracy and scale are critical.

Content trapped in one team, tool, or format where no one else can reuse or govern it. Silos are where duplication, drift, and out-of-date information quietly multiply.

Writing a piece of content once and publishing it to many outputs and formats from that one source. The opposite of maintaining separate copies for print, web, and help.

Writing in standalone topics that each cover one idea, instead of long linear documents. Topics are easy to reuse, reorder, and assemble into different deliverables.

A text format that wraps content in tags to mark its meaning and structure. It makes content machine-readable, though it can be heavy to author by hand.

An XML standard for structured, topic-based technical content. Powerful and widely used, but it carries real complexity and overhead that not every team needs to take on.

Content marked to appear only in certain outputs, such as one product variant, region, or audience. It lets a single source serve many versions without separate documents.

A placeholder that holds a value, such as a product name or version number, filled in at publish time. Change it once and every output updates, with no find-and-replace.

Data about content: its type, status, version, language, owner, and more. Metadata is what makes content findable, governable, and usable by AI systems.

The rules, roles, and workflows that decide who can create, change, approve, and publish content. Governance is what makes content trustworthy at scale.

A defined process where content is checked and formally signed off before it can be published. It creates the accountability that regulated industries and reliable AI both depend on.

A governance control that stops content reaching an output until it has been approved. The gate is what ensures only trusted content flows downstream, including into AI systems.

A recorded history of who changed what and when. Audit trails are essential for compliance and for proving that published content is the approved version.

Keeping track of changes to content over time so you always know the current approved version and can see what changed. It prevents the chaos of competing copies.

A version of the same core content adapted for a specific product, market, or audience, produced from one source rather than maintained separately.

The automated path content takes from its single source to a finished output, applying formatting, conditions, and variables along the way.

The shape content is published into, such as Word, HTML, PDF, or structured JSON. One source can produce many output formats without rewriting anything.

Handling content in multiple languages from a single source, so updates flow to every language and nothing falls out of sync. Reuse means only changed content needs retranslating.

Content that must meet regulatory or safety requirements and prove it is current and approved. Structure, governance, and audit trails are what make compliance defensible.

Being able to follow a piece of content back to its source, its approvals, and its changes. Traceability is what turns a claim into something you can stand behind.

Author-it's structured JSON publishing output, introduced in 2026.R1. It turns governed, single-source content into clean, metadata-rich JSON built for ingestion by LLMs, RAG pipelines, and AI agents.

The structured, governed, single-source content layer that AI systems need to answer accurately. AI is only as reliable as the content it draws on, so the foundation comes first.

A Large Language Model. The kind of AI that generates text by predicting likely words. It is fluent but has no built-in source of truth, so the quality of its answers depends on the content it is given.

Retrieval-Augmented Generation. A method where an AI retrieves relevant source content first, then uses it to answer, instead of relying only on what it memorised. Clean, structured content is what makes retrieval work.

Tying an AI's answer to specific, retrievable source content so the response can be trusted and traced, rather than invented. Structured content is what makes grounding reliable.

When an AI produces confident but false or unsupported information. The most effective defence is grounding the model in accurate, governed source content.

The process of feeding content into an AI system so it can be retrieved and used. Messy, unstructured content breaks at this stage; clean, structured content flows straight through.

Splitting content into smaller pieces so an AI can retrieve the right part rather than a whole document. Content already written as components is, in effect, pre-chunked.

A numerical representation of a piece of content that lets an AI find things by meaning rather than exact words. Good structure and metadata produce more accurate matches.

A store of embeddings that an AI searches to find the most relevant content for a question. It is the retrieval layer behind most RAG systems.

Search that matches on meaning rather than exact keywords, so a question finds the right answer even when the words differ. It relies on well-structured, well-described content.

The record of where a piece of content came from, who approved it, and when. Provenance is what lets an AI answer cite a trustworthy source rather than an anonymous fragment.

A reference an AI gives back to the source it used for an answer. Citations are only as trustworthy as the content behind them, which is why governed sources matter.

The body of content an AI system draws on to answer questions. A governed, structured knowledge base is the difference between helpful answers and confident guesses.

Answer Engine Optimisation. Structuring content so AI answer engines can extract, trust, and cite it. Clear definitions, question-and-answer formats, and schema all help.

Content delivered as JSON with clear fields and labels, so machines can read it precisely. It is the format AI pipelines prefer because nothing has to be guessed or scraped.

Pulling content off published pages or files to feed an AI, often losing structure, metadata, and source in the process. It is the fragile alternative to ingesting structured content directly.

A proposed standard file that tells AI systems how to find and use a site's most important content. Part of a broader shift toward making content deliberately AI-readable.

Where good content goes to die. Information locked inside a flat PDF that no one can reuse, update, or feed to anything, until someone retypes it by hand.

The game you play when the same paragraph lives in fourteen documents and you have no idea which one is current. Nobody wins.

The act of digging through old files, shared drives, and inboxes to work out which version of a document is the real one. Hours lost, truth rarely found.

All the content you keep meaning to fix, update, or consolidate but never quite get to. Like technical debt, it compounds quietly until something breaks.

What happens when copies of the same information slowly diverge over time, until no two versions agree and no one can say which is correct.

The state of having so many near-identical files (final, final_v2, final_USE_THIS) that the true version is anyone's guess. Served daily, usually right before a deadline.