Frequently asked questions
How is Citara different from ChatGPT?+
ChatGPT is free-form conversation. Citara is a structured environment: source registry, [Sxx] citations, chapter plan, quality checks, and DOCX export. AI writes from your registered sources.
What about Notion AI or Copilot in Word?+
Those are assistants inside a generic editor. Citara separates content (Markdown) and layout (export), with systematic traceability — built for long, auditable dossiers.
Do I need to supply API keys (OpenAI, etc.)?+
Not at this stage. The AI model is operated by the Citara platform—you do not configure a provider or keys. You work on the dossier; we handle model calls within your workspace and privacy settings.
Which ingest formats?+
PDF and DOCX are supported via the ingest pipeline (Docling). Other formats may be added per deployment.
How do [S01] citations work?+
Each registry source gets an ID (S01, S02…). Factual claims in the document reference them inline. The review job checks consistency before export.
Is data used to train a model?+
Not in our operating model: your content feeds your dossiers, not third-party model training. Contractual terms specify commitments per deployment.
Where is data hosted?+
Shared cloud in Switzerland (CH datacenter, confirm in contract). Enterprise: private cloud or on-premise in Switzerland.
Can LLM logs be disabled?+
Yes. Workspace privacy mode (log_privacy) prevents persisting request/response text; token usage metadata may still be logged.
On-premise deployment?+
Available on Enterprise with dedicated support.
Generated content languages?+
FR, DE, EN, IT per your briefs and sources — suited to multilingual dossiers in Switzerland. The model follows document scoping.
How does the pilot work?+
3–6 months on a limited scope (one dossier type, one team). Goal: validate workflow and metrics before wider rollout.
Per-dossier pricing?+
Option for IT services and RFP teams: per-dossier rather than per-seat billing, on quote.