What does the high-level architecture look like?

At a high level, Dynamically Generated Experiences (DGE) in Sitefinity CMS follows a simple flow: intent → retrieve → generate → render.

A visitor expresses intent through an input interaction, or Sitefinity CMS can infer intent from Sitefinity Insight CDP signals. DGE then retrieves the most relevant context from approved, published Sitefinity content using retrieval-augmented generation (RAG). The content is chunked, embedded and indexed, and relevant passages are retrieved through vector matching and scoring.

An LLM assembles the experience from the retrieved passages, using the packaged context to follow the configured tone, voice, templates and rules. The generated content is then rendered inside the layout selected by editors, so brand, accessibility and presentation patterns remain consistent.

• Intent layer: Captures visitor intent through an input widget or inferred Sitefinity Insight signals.
• Retrieval layer: Uses RAG to find relevant passages from published Sitefinity content.
• Generation layer: Uses an LLM to compose the experience from grounded content.
• Presentation layer: Renders the generated output through editor-controlled widgets, layouts and templates.
• Governance layer: Keeps drafts, deleted items, unpublished content and restricted content out of indexing.

Architecturally, DGE is delivered as widgets that can be placed on any Sitefinity page. Teams can use it for a single high-impact section, multiple dynamic sections or a fully DGE-driven page. It also supports caching, default queries, observability and safe fallbacks so that, if generation times out, visitors see the default experience instead.