Every piece of content is a mesh member from the moment it exists.
The Writer generates four content types, each with its own structure. Every generation call knows the site's voice, the active strategy, and the cluster the content belongs to.
Each content type has its own structure.
The Writer does not produce one shape of content. It produces the right shape for each role in the mesh.
Blog posts
Article structure — introduction, sections, tips, and FAQ.
Pages
Service or informational pages, scoped to a cluster.
Products
Descriptions with features, specifications, pros and cons.
Taxonomy terms
Full definition pages — H1, 500-1500 words, FAQ block, related terms, and meta.
Content that sounds like it belongs to the site.
Brand guardrails are injected into every AI call — not like a machine that has never read the site.
The site's voice profile
Tone, structure, and vocabulary, applied to every generation.
The active strategy
Content is produced against the current SAG plan.
The cluster it belongs to
Each piece knows its role in the mesh before it is written.
The content template for its type
The structure each content type requires, applied automatically.
Image prompts are extracted from content automatically.
Images are generated and attached to content, ready for publishing — not a separate step the operator runs.
| Provider | Role |
|---|---|
| DALL-E 3 | Generative image provider |
| Runware | Generative image provider |
| Bria AI | Generative image provider |
- Prompts extracted from the content itselfAutomatic
- Generated by one of three providersGenerate
- Attached to content ready for publishingAttached
Content from a blank prompt
Output that has never read the site — no voice, no strategy, no place in any structure.
- No voice profile applied
- No connection to a strategy
- No cluster assignment
- No content-type structure
Content from full context
Voice, strategy, cluster, and content template on every call — a mesh member from the moment it exists.
Four content types. Three image providers. One coherent voice.
Where AI content usually falls down.
Content that has not read the site does not belong to it.
| Anti-pattern | Why it fails | The Writer approach |
|---|---|---|
| Generic AI output | Reads like a machine that has never seen the site | Voice profile injected into every generation call |
| One structure for everything | Products, posts, and taxonomy terms need different shapes | Four content types, each with its own template |
| Content written in isolation | No cluster, no role, no place in the structure | Every piece assigned to its cluster before it is written |
| Images sourced separately | Slow, manual, disconnected from the content | Prompts extracted from content; images generated and attached |
From context to attached content.
Every generation follows the same path — context first, structure second, images attached last.
- 01Load contextVoice, strategy, cluster, and content template assembled.
- 02GenerateContent produced to the structure its type requires.
- 03Extract image promptsPrompts pulled from the content automatically.
- 04Attach imagesGenerated by three providers, attached ready for publishing.
Content that belongs to the mesh.
Connect a URL and the Writer generates against your voice, your strategy, and your clusters.