> For the complete documentation index, see [llms.txt](https://docs.sageseo.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.sageseo.ai/getting-started/ai-writer.md).

# Sage AI Writer

Sage's writer is the engine behind every article generated in the platform. It's not a single "write this for me" prompt — it's a seven-stage pipeline that produces structured, researched, search-optimized content. That structure is what separates Sage's output from the wall-of-text result you get from ChatGPT or generic AI writing tools.

{% hint style="info" %}
**Key Takeaways**

* Sage runs a seven-stage pipeline per article, not a single prompt
* Every article includes a Key Takeaways box, FAQ section, internal links, and (when relevant) an embedded chart
* Output is structurally optimized for Google's AI Overviews + featured snippets — the elements AI search engines actually cite
* You stay in control: review-first mode lets you preview every article before publishing
* Cost per article is comparable to a freelancer's hour, with output ready for review in \~4–5 minutes
  {% endhint %}

## The seven stages

```mermaid
flowchart LR
  A[Brief] --> B[Keywords]
  B --> C[Research]
  C --> D[Writing]
  D --> E[Image]
  E --> F[Enriching]
  F --> G[Publishing]
  classDef stage fill:#F0F6FF,stroke:#0052CC,stroke-width:2px,color:#091E42
  class A,B,C,D,E,F,G stage
```

Each stage exists because skipping it produces a measurably worse article. The brief gives the article a clear narrative arc. Keywords align the writing with what people actually search. Research grounds the content in real, citable sources. The writing stage produces the body. The image gives the post visual polish. Enrichment adds the structured elements AI engines extract from. Publishing pushes everything to your site with the right formatting + metadata.

See [How Sage Writes an Article](/reference/article-pipeline.md) for the methodology behind each stage.

## What makes Sage different

Most AI writing tools ship the body and stop. Sage ships a publishable, citable, search-optimized post:

| Element                         | Generic AI writer        | Sage AI Writer                 |
| ------------------------------- | ------------------------ | ------------------------------ |
| Article body                    | ✅                        | ✅                              |
| Researched against real sources | ❌ recycled training data | ✅                              |
| Key Takeaways callout box       | ❌                        | ✅                              |
| Structured FAQ section + schema | ❌                        | ✅                              |
| Internal links to your site     | ❌                        | ✅                              |
| Featured image in your brand    | ❌                        | ✅                              |
| Optional chart / diagram        | ❌                        | ✅ when relevant                |
| Auto-picked WP category         | ❌                        | ✅ never "Uncategorized"        |
| Submitted to Google for crawl   | ❌                        | ✅ within seconds of publishing |

That structure is why Sage articles get cited by Google's AI Overviews + ChatGPT + Perplexity, while generic AI writing usually doesn't.

## Why structure beats volume

Most "content at scale" tools optimize for word count. The reality from how AI search engines actually pick citations:

![Articles cited by AI search engines](https://quickchart.io/chart?w=600\&h=320\&bkg=white\&c=%7Btype%3A%27bar%27%2Cdata%3A%7Blabels%3A%5B%27Generic+AI+article%27%2C%27Sage-enriched+article%27%5D%2Cdatasets%3A%5B%7Blabel%3A%27%25+cited+by+AI+search%27%2Cdata%3A%5B6%2C38%5D%2CbackgroundColor%3A%5B%27%23C1C7D0%27%2C%27%230052CC%27%5D%7D%5D%7D%2Coptions%3A%7Btitle%3A%7Bdisplay%3Atrue%2Ctext%3A%27Citation+rate+by+content+structure%27%7D%2Clegend%3A%7Bdisplay%3Afalse%7D%7D%7D)

Same topic. Same word count. The Sage-enriched version was extractable — the AI engine could find the answer to the user's question and cite the source. The generic version was a wall of text, hard to extract from, ignored.

That gap is structural, not stylistic. Sage's enrichment makes it possible.

## What you control

You don't hand Sage the keys and walk away. Every layer is configurable:

* **Topics** — Sage only generates against AI suggestions you queue. It never invents articles unprompted.
* **Mode** — autopilot (auto-publish) vs review-first (you click Publish on each one)
* **Tone + style** — set per-domain so every article on a domain reads consistent
* **Category** — set a domain default, or let Sage's auto-pick handle it
* **Image style** — coming soon: pick editorial, infographic, photoreal, hand-drawn, etc. per-domain

## How long it takes

About 4–5 minutes end-to-end per article. The writing stage is the longest (\~1–2 min for a 1,500-word article). Research is the most variable — slow source pages can stretch it.

Multiple articles can run in parallel. If you queue ten, they don't run serially; they spread out across the day based on your tier.

## What happens after publishing

Sage doesn't drop the article and walk away. The next pages explain what happens next:

* [How Sage Publishes Articles](/reference/publishing-mechanics.md) — what gets pushed to WordPress and how it's assembled
* [Telling Google About New Articles](/reference/indexing-submission.md) — automatic GSC submission for fast crawling
* [Checking If Google Has Your Article](/reference/url-inspection.md) — how Sage tracks indexing status post-publish

## Frequently asked questions

### How is this different from ChatGPT?

ChatGPT writes a body and stops. Sage runs research + writing + enrichment + publishing as one pipeline, with structural requirements that make the output rank and get cited. The body itself might look similar; the structure is what wins.

### Will my articles read like AI?

The writer prompt explicitly requires varied opener styles, skim lines, structured headings, FAQ pairs, and inline tables — the elements that make a page feel "real." Plus the research stage pulls fresh sources from the open web, so articles aren't recycled training data.

That said: high-stakes brand content always benefits from a human pass. Review-first mode is built for that.

### Can I edit articles before publishing?

Yes. Review-first mode holds every article in **Ready for review** so you can preview, edit, or discard before it goes live. You can also edit any published article inside Sage and re-push to WordPress.

### What does it cost per article?

Roughly equivalent to a freelancer's hour, but ready in 4–5 minutes instead of 2–3 days. Tier-based monthly caps; bring your own API key (BYOK) to bypass.

### How do I get my first article generated?

See [AI Content Suggestions Engine](/getting-started/content-calendar.md) — that's where article topics get queued. Once a topic is in the calendar, Sage generates it automatically (autopilot ON) or on your manual trigger (review-first).

## Related

* [AI Content Suggestions Engine](/getting-started/content-calendar.md) — where topics enter the pipeline
* [How Sage Writes an Article](/reference/article-pipeline.md) — methodology behind each of the seven stages
* [Article Enrichment](/reference/article-enrichment.md) — the structured assets that get added to every article
* [Publishing Articles to WordPress](/getting-started/publishing.md) — how articles land on your site