# Review & scheduling How content gets planned, written, approved, and published — and where you stay in the loop. ## What language a blog is written in Each blog publishes in one language, chosen on the blog's settings page. It is not a label stuck on top of English output: the planner, the writer and the editor are all prompted **in that language**, from a prompt pack written in it, so an article comes out as something written in Romanian rather than something translated into it. The topic queue is planned in it too — the phrases a reader actually types into a search box are not what translating an English topic list produces. The setting is also what the outside world is told: - every server-rendered page declares it as ``, - the [delivery API](/docs/delivery-api) defaults `?locale=` to it, - the featured image's `alt` text — which a screen reader announces mid-article — is written in it. | Language | Tag | |---|---| | English | `en` | | Deutsch (German) | `de` | | Español (Spanish) | `es` | | Français (French) | `fr` | | Italiano (Italian) | `it` | | Nederlands (Dutch) | `nl` | | Português (Portuguese) | `pt` | | Română (Romanian) | `ro` | A site can run several blogs in different languages — one blog at `/blog`, one at `/blog/ro` — and each is planned, written and served in its own. :::callout{severity=warning title="Changing it moves the articles you already have"} Articles already published move to the new language along with the blog, so the archive keeps serving and none of its URLs break. What they *say* stays as written until you regenerate or edit them. Topics still queued from the old language are kept but no longer used — a brief carries the language it was planned in, and generation only spends one matching the blog. They stay in the queue, marked, in case you move back. Plan topics again to fill it in the new language. ::: Need a language that is not on the list? Ask for it. A language here is one prompt pack, not a change to the product. ## The review queue Every generated post lands in **pending review**. The post page shows: - the full rendered preview (exactly the components readers will see), - the **backlinks** the article carries — anchor text and target URL for each placement, - the **generation trace** — every pipeline step with model, token counts, latency, and status. **Approve & publish** makes the post live on your delivery surface within a minute (CDN caches revalidate on a short window). Nothing ships without a click unless you opt in to auto-publish. :::callout{severity=warning title="Auto-publish is per blog, and off by default"} Toggle it in the blog's schedule settings when you trust the output. Scheduled generations then go straight to published. You can turn it off at any time; posts already published stay published. ::: ## The topic queue **Plan topics** asks the planner for article briefs grounded in your site profile — title, target keyword, secondary keywords, and an outline. Briefs arrive as *proposed*; you approve or reject them. Approved briefs are consumed first — both by the scheduler and by the **Generate now** button — before any ad-hoc topic planning happens. Keeping a handful of approved briefs in the queue means you always know what's coming next. ## Schedules Each blog has one schedule: | Setting | Values | |---|---| | Cadence | weekly · biweekly · monthly | | Posts per period | 1–10 | | Publish hour | 0–23, in the schedule's timezone | | Timezone | any IANA zone, e.g. `Europe/Bucharest` | | Paused | stops the schedule without losing its settings | When a run is due, the scheduler creates the post (consuming an approved brief if one exists), generates it, and routes it to review — or straight to published with auto-publish on. The next-run time is always visible on the blog page. ## Quotas and failure behavior - Generation consumes one post credit at creation time; your plan's monthly quota and usage meter live on the site page ([plans & quotas](/docs/billing)). - When quota runs out, the schedule **pauses itself** instead of failing silently or billing you by surprise. - If a generation step fails, the post is marked *failed* with the error in its trace — a failed run doesn't publish anything half-finished.