AI / LLM Transparency

1. What We Send

• Snippets of public pages you asked us to analyze (title, extracted text, structural signals).
• System/user prompts to generate drafts (e.g., llms.txt) or to run Measured citation probes.
• We do not send passwords, payment cards, or content from authenticated areas of your site.

2. Purposes

Generation of operational artifacts and sampling of Measured citation share on the engines enabled by the plan. Estimated scores remain based on structural crawling without LLM probes.

3. Providers and Training

Active providers are listed in the DPA / Annex A (OpenAI, Anthropic, Perplexity, Google Gemini, xAI if configured). Centropic does not use your content to train its own models. Provider use of data for training depends on their enterprise/API policies: we use API endpoints and available settings to limit retention/training where supported. For binding details, see each provider’s privacy policy.

4. Logs and Retention

We retain job metadata, token charges, and results necessary to provide the service. Artifact packs may remain in object storage with configured retention (typically 90 days for technical artifacts). You can export or delete the account from Settings.

5. Limitations and Non-Guarantees

Models may produce incomplete or inaccurate outputs. Estimated ≠ Measured. Centropic does not guarantee rankings or mentions in third-party UIs. Publication of artifacts and compliance remain your responsibility.