Frequently asked questions
Straight answers on pipeline depth, retention, BYOK, and how we treat citations and contradictions.
How is this different from ChatGPT Deep Research or Claude Research?
ResearchOne exposes a named ten-stage pipeline with a Skeptic outside the single-model draft, per-claim citation bind, and explicit source-corroboration tiers. The comparison is structural auditability versus a single chat-shaped pass.
What does the Skeptic agent actually do?
It receives the draft report and argues against it: searches for counter-sources, stress-tests claims, and returns counter-claims with contradictions preserved for the Citation Bind stage.
What happens when two high-tier sources disagree?
Both survive into the report with tier tags and a named contradiction block — no silent consensus rewrite.
Can ResearchOne forge or hallucinate a citation?
Claims without a verified source span do not ship as cited. The Verifier gate blocks or downgrades them first.
How do Living updates work — will my old report change under me?
Each update is a new version with a diff trail. You pin, compare, or roll back; prior agreed versions remain addressable.
What's the difference between BYOK and the Sovereign tier?
BYOK brings your keys into our orchestration plane. Sovereign runs the stack in your tenancy with contract isolation and custom retention.
Can I export a report and cite it like a static PDF?
Exports include citations and version identifiers so a frozen PDF remains traceable to the corroboration state at export time.