How Sage Writes an Article

How Sage turns a topic idea into a finished, publishable article — and why each step exists.

Sage's article pipeline isn't one giant "write this for me" prompt. It's seven distinct steps, each with a specific job. The result is a structured, researched, polished article — not a wall of AI text.

The seven steps

1. Brief — Sage expands your topic idea into a structured plan: target word count, tone, audience, key questions the article should answer, suggested outline. This is the equivalent of an editorial assistant turning a one-line pitch into a brief.

2. Keywords — primary keyword + 2–5 secondary keywords. Pulled from your existing GSC rankings, competitive analysis, and topical relevance.

3. Research — Sage's writer fetches real sources — recent studies, statistics, expert articles — related to the topic. This is what separates Sage from a generic AI writer. Generic AI recycles training-data content; Sage incorporates fresh, citable material.

4. Writing — the article itself, built against the brief + keywords + research. Sage's writer prompt requires specific structural elements: Key Takeaways box near the top, at least one table, at least one ordered + unordered list, FAQ section, one-sentence direct answers under question headings, and "skim lines" opening each section.

5. Image — an editorial illustration generated specifically for this article, in your brand colors. The image is selected via a two-step prompt: first identifies the most concrete photographable concept from the article, then renders it in the editorial style of NYT/Atlantic opinion sections — not generic stock-vector SaaS imagery.

6. Enriching — adds Key Takeaways, FAQs, charts, internal links. See Article Enrichment for what each piece does.

7. Publishing — if autopilot is on, the finished article goes straight to your WordPress site. If review-first, it lands in ready_for_review for your approval.

Why each step exists

Step	What goes wrong without it
Brief	Article wanders — no clear narrative, no purpose, vague tone
Keywords	Article ranks for nothing — no consistent topical signal to Google
Research	Article reads like recycled AI text — same generic facts as everyone else
Writing	(this is the obvious one — no article)
Image	Post looks unfinished, gets ignored on social, fails featured-image SEO checks
Enriching	Article ranks but doesn't get cited by AI Overviews or featured snippets
Publishing	Manual hand-off bottleneck — defeats the point of automation

What we DON'T do

• No "one big prompt that does everything" — those produce wall-of-text, generic, structurally-bad output. The seven-step pipeline produces fundamentally different quality.

• No copying competitors verbatim — research pulls factual citations + recent context, not phrasing. The writing step generates original prose against the brief.

• No keyword stuffing — keywords inform the brief, the H2 structure, and the natural placement of phrases. They are not artificially repeated.

Total time per article

Roughly 4–5 minutes end-to-end. Research is the most variable step — slow web responses can stretch it. Writing is the longest single stage (1–2 minutes for a 1,500-word article).

Tier-aware cost

Each article costs roughly $1.50–2.00 in AI inference. Tier caps in Sage's pricing exist because high-volume usage at that cost adds up. Customers running serious volume often switch to BYOK (bring your own API key) to bypass the caps.

Related

• Autopilot — when the pipeline runs automatically

• Article Enrichment — what happens in step 6

• How Sage Publishes Articles — what happens in step 7