Most CCMSs publish documents. Author-it publishes data your AI can use.

A traditional CCMS manages structured authoring and typically publishes to PDF, HTML, and DOCX - formats designed for human readers. An AI-ready CCMS extends that foundation to publish structured data formats that AI systems can consume directly, with content type information, metadata, and hierarchy preserved in the output. Author-it's AION output is the example: it publishes structured JSON from the same content library that produces PDF and HTML, providing AI systems with typed, metadata-annotated, component-level data rather than formatted documents.

You can - but the quality of AI output will be limited. When an AI or RAG system ingests a PDF, it receives stripped text. Content type information is lost (the AI can't distinguish a safety warning from a feature description). Hierarchy is lost (the AI doesn't know a section belongs to a sub-procedure). Metadata is lost - no authorship, no modification date, no version. Chunks are created at arbitrary paragraph or page boundaries rather than meaningful topic boundaries. All of this reduces retrieval precision and increases the likelihood of inaccurate, decontextualised answers.

AION includes the following on every published topic: the template name and type (which defines the content type - task, warning, concept, and so on), the topic ID and book ID, the library folder path, the last-modified timestamp and the name of the author who made the last change, the topic description, and all variable values resolved at publish time (such as product name, region, or version). This metadata travels with every content chunk, giving AI systems the context they need to deliver accurate, citable answers.

Retrieval-augmented generation (RAG) works by retrieving the most relevant text chunks from a knowledge base and passing them to an LLM to generate an answer. When content is chunked from a PDF at arbitrary paragraph or page boundaries, the retrieved chunks may contain unrelated content, split mid-procedure, or lose the context that gives them meaning. AION chunks content at the topic level - each chunk is a discrete, typed, metadata-annotated unit authored as a single coherent piece. These chunks produce more precise vector embeddings, which produce more accurate retrieval, which produces better LLM answers.

AION is a publishing output, not a live scrape or continuous sync. Content must go through Author-it's standard publishing workflow to reach AION output. In organisations using Author-it's Review and Approvals module, only approved content can be published. Draft, in-review, or expired content stays within the authoring environment and does not appear in AION output until it has been approved and published. This means the same governance that controls what goes into a PDF controls what goes into your AI.

In a document-centric approach, content is created and managed as whole documents - Word files, PDFs, web pages. When those documents are fed to an AI system, the AI receives the final rendered output: text without structure, context, or type information. In a component-centric approach, as used by Author-it, content is authored as discrete topics - individual units with defined types and metadata. These components are assembled into documents for human readers and published as structured data for machines. The AI receives components, not documents: smaller, typed, metadata-annotated units that produce more precise retrieval and more reliable answers.

No. Any book already in the Author-it Library can be published to AION without restructuring. A Library Administrator creates an AION publishing profile once - selecting Resolved XML format and adding the JSON conversion step - and from then on, any author can publish to AION by selecting that profile. Organisations that want to improve AI output quality can do so by improving their library design (better topic typing, richer metadata, cleaner variable usage), but this is optimisation, not a prerequisite. Existing structured content works as-is.

Most CCMS platforms currently export to PDF, HTML, DITA XML, or proprietary formats. These outputs are designed for human-facing delivery and require additional processing - scrapers, parsers, or manual conversion pipelines - before they can be used in a RAG or AI workflow. AION is Author-it's dedicated AI output format: structured JSON produced directly from the publishing workflow, with metadata and hierarchy built in. No additional pipeline is required.

Content hierarchy tells the AI how individual pieces of knowledge relate to each other - which topic belongs to which procedure, which procedure belongs to which product area, which product area belongs to which product line. Without hierarchy, an AI retrieving a chunk about torque settings has no way of knowing whether it applies to the ProSeries 300 or the ProSeries 400, or whether it is part of an installation procedure or a maintenance checklist. AION preserves the full book-to-sub-book-to-topic hierarchy, giving the AI the organisational context it needs to distinguish between similar content and answer questions precisely.

Governed AI output means that the content your AI systems receive has passed through the same review, approval, and publishing controls as your other outputs. In Author-it, this means content that has not been approved does not reach AION. Content published to AION carries modification history, authorship, and content type - so answers produced from it can be audited back to a specific approved component. For regulated industries where hallucinations are a compliance event, governed AI output is the difference between an AI tool you can trust and one you cannot.