Skip to content

Glossary

What is grounding?

Grounding ties an AI output to chosen context or evidence. It can make an answer easier to support, but both the sources and the claims still need checking.

Updated 21 Aug 2026

01

Where this one gets misread

Grounded is used as a synonym for safe. It means the output was produced with reference to supplied context, which says nothing about whether that context was the right context or whether the claim follows from it.

02

Questions to ask

Ask for a worked example on your own material, and the evidence needed to reproduce it.

  • What was retrieved, and can you see it?
  • Does every claim carry support, or only the answer as a whole?
  • What happens when two sources disagree?
  • What does it do when nothing relevant was found?
03

How Marella uses the term

We use “Grounding” only where a product mechanism or an evaluation method backs it up, and we say when the behaviour depends on how a deployment is configured.

  • Backed by a product mechanism or an evaluation method
  • Deployment differences flagged

What this page does not prove

  1. B1A definition is not a claim about how the product performs.
  2. B2Vendor implementations vary.
  3. B3Test the term against a representative workflow.

Test the claim on your documents

Pick a real piece of work, agree what a good answer looks like, then go through the results together.