Citerva is in early access. This first release covers the full loop: describe your brand, review the questions we will ask, run a scan, and read the results.
What works today
Projects and prompt sets. You describe your product, your competitors and the topics you care about. Citerva generates a set of questions from that description and shows them to you before anything is sent anywhere. You can edit, add, reject or regenerate them. Generating and editing prompts is free — a question only reaches an AI provider when you approve it and run a scan.
Scans across configured providers. A scan asks each approved question and stores the raw answer exactly as it came back. Responses are append-only: metrics are derived from them and can be rebuilt, but the original is never overwritten.
Mention rate, share of voice and citations. Every figure is shown with the denominator it was computed from, and that denominator counts only runs that actually succeeded. A provider that failed or was skipped is reported as failed or skipped — never silently counted as "did not mention you".
What Citerva does not claim
A scan is a sample of a system that is not deterministic. Ask the same question twice and you may get two different answers, so a single scan is evidence, not a verdict — which is why the answer behind every number is one click away.
Nothing here predicts or guarantees an outcome. Citerva can show you which questions you are absent from and which sources are being cited instead. It cannot promise that any particular change will make a model recommend you.
Known limits
Provider coverage depends on which credentials are configured for your installation. The dashboard lists exactly which providers ran for each scan, so coverage is always visible rather than assumed.