CiteFlow drafts AEO-ready articles by combining your knowledge base, target keyword and audit signals into a single brief.
Article generation
Once a topic comes off the queue, CiteFlow generates a full article: title, slug, meta description, body, FAQ, schema markup, suggested keywords, and a hero image. This page covers how it works, what you can edit, and how to ship it.
How does the generation pipeline work?
- A topic is picked from the queue.
- Claude Sonnet generates a 1,500–2,000 word article using your approved knowledge base as context.
- The article is structured AEO-first: an answer-first opening paragraph, scannable subheadings (often phrased as questions), and a dedicated FAQ section at the end.
- JSON-LD schema is generated:
Articlealways,FAQPagewhen FAQ items are present. - UK English is enforced at the prompt and re-checked post-generation.
- A hero image is generated in parallel (see below).
- The finished article lands in your dashboard as a draft.
Typical generation time is 60–120 seconds end-to-end.
What is the structure of a generated article?
Every generated article has:
- Title, H1, optimised for the target search intent.
- Slug, URL-friendly, derived from the title. Editable before publish.
- Meta description, 150–160 characters, distinct from the opening paragraph.
- Body, markdown and pre-rendered HTML. Either is exported to your CMS depending on the adapter.
- FAQ items, extracted into a structured list for the
FAQPageschema and for CMS adapters that surface them separately. - Schema JSON-LD,
Articleand (when applicable)FAQPage. - Target keywords, five suggestions surfaced for review. Not injected into the article; they're for your tracking decisions.
- Hero image URL and alt text.
How are hero images created?
Hero images are generated by Nano Banana Pro
(gemini-3-pro-image-preview), prompted to produce a 16:9
editorial-style image appropriate to your industry and the article's
topic.
- Soft-fail behaviour, if image generation fails (rate limits, policy filter, transient API issue), the article still completes without a hero. You'll see a "no image" placeholder and a Regenerate image button.
- Regenerate, click to try again with a fresh prompt seed. Each attempt counts towards your image cost.
- Upload replacement, accepts PNG, JPEG, or WebP, max 8 MB. The uploaded image becomes the hero, replacing any generated one.
- Alt text, editable in the article editor. Keep it descriptive (it ships with the article to your CMS).
What do the article statuses mean?
An article moves through:
- draft, just generated. Not visible to your audience until you approve and publish.
- approved, you've reviewed and signed off. If auto-publish is enabled (see scheduling), publishing triggers immediately.
- published, sent to one or more configured publishing endpoints.
- archived, removed from active lists. Reversible.
How do you edit an article?
The article editor lets you adjust:
- Title, slug, meta description.
- Body markdown.
- Each FAQ item.
- Hero image and alt text.
- Suggested keywords (add or remove from tracking).
Schema JSON-LD is regenerated automatically from the edited content at approval time.
How do you approve and publish an article?
- Open the article from
/dashboard/content. - Review and edit.
- Click Approve.
- If auto-publish is on and at least one endpoint is configured, the article ships immediately. Otherwise, click Publish and pick the endpoint(s).
See publishing overview for endpoint setup, and auto-scheduling for the auto-publish toggle.
What are the tier limits?
| Tier | Articles per month |
|---|---|
| Trial | 2 articles total |
| Standard | 10 per month |
| Enhanced | 30 per month |
| Marketplace | Unlimited (soft fair-use cap) |
The dashboard shows your remaining allowance. Once exhausted, generation pauses until the next billing period or until you upgrade.
How is article cost calculated?
Article generation cost (Claude Sonnet + Nano Banana Pro) is included in your subscription. There are no per-article add-on charges. We monitor cost-per-tenant and the Marketplace soft cap exists so a single tenant generating tens of thousands of articles a month can be individually rate-limited if needed.
Related
- Topic queue, what feeds the generator
- Knowledge base, the brand context every article uses
- Auto-scheduling, cadence and auto-publish
- Publishing overview, where articles go after approval
References
- Schema.org Article specification, Schema.org
- Schema.org FAQPage specification, Schema.org
- Google Search Central: structured data guidelines, Google
Related
- Topic queueSources, reorder, skip, bulk add. How a topic becomes the next generated article.
- Knowledge baseAuto-synthesised from your audit. Review, edit and approve before content generation reads it.
- Auto-schedulingCadence, timezone, hour-only granularity, notification email, auto-publish and trial-expiry pause.
- OverviewPick the adapter that matches your stack. All adapters create drafts so your team reviews before publishing.