Grounding
Also called Source grounding, Retrieval grounding
Grounding means an AI tool answers from a given set of sources rather than from general training data, and can point to the passage it used.
Why it matters in a workup
A grounded answer is easier to check. You can open the source and see whether it says what the answer claims.
Grounding reduces errors but does not remove them. A tool can still misread a source or combine two facts wrongly.
Put it into practice
- Ask which source each answer comes from
- Open the source and check the passage
- Treat an answer with no source as unverified
Frequently asked questions
Does grounding mean the answer is correct?
No. It means you can check it. You still read the source.
General guidance on clinical documentation, not advice about any individual patient and not legal advice.
Glossary
Look up related terms
- HallucinationA hallucination is a statement an AI tool produces that sounds plausible but is not supported by its sources or by fact, such as an invented result.
- Source documentA source document is the original record a fact comes from, such as a lab report, an ECG report or a letter. Checking it is how you verify a statement.