SciGroundLimits and conditions

What SciGround proves.And what it doesn’t.

SciGround makes strong, specific guarantees. This page states their boundaries in one place; each capability page links to its own detailed limits.

What “verified” means

Specified properties under stated assumptions: A verified result establishes that the data, calculation, and claim are connected to the same research target and satisfy defined checks under the stated assumptions.
Not scientific truth: SciGround does not prove biological conclusions, the correctness of measurements, or the truth of a hypothesis.
Statistics remain statistics: Statistical significance does not establish truth, and error rates describe procedures, not the probability that a particular claim is correct.
Your declarations stay declarations: Independence, randomization, blinding, and how the study was conducted are recorded as your declarations. Answering a question or clicking confirm does not prove them.

Results and Records it issues: SciGround verifies the results and Records it issues. Manuscript findings are labeled rule-checked, AI-assessed, or not checkable. It does not check every answer your AI writes.
Supported workflows: Analyses outside the supported workflows are refused, not approximated. See Supported analyses.
History that exists: SciGround cannot create pre-registration after the fact, and it does not detect data registered again as a new study.

Life-science research data: SciGround is designed for life-science research data that are not identifiable and not clinical. It is not for clinical use.
Through your AI: SciGround is used through MCP and the API from supported AI environments. It is not a standalone statistics application.
Scale: Workflows are validated on research-scale datasets; no large-scale performance guarantee is made.
Your responsibility: Study conduct, ethics approval, interpretation, and the final manuscript remain the researchers’ responsibility.