Glossary term 17 of 33
Grounding
Anchoring a generated answer in retrieved source material rather than in the model's training data alone.
What does Grounding mean?
Anchoring a generated answer in retrieved source material rather than in the model's training data alone. Grounded answers are the ones that cite, which makes them the ones worth competing for.
Quoted without alteration from the glossary of ECHO: How to Make AI Recommend Your Brand.
Which ECHO pillar does Grounding belong to?
Grounding belongs to Hooks, pillar 3 of the ECHO framework. Does it find your content when the question is asked?
The third ECHO pillar. Structured, extractable, self-contained passages that an AI can lift out and attribute without losing meaning. The strongest hooks are ones only you can create: original data, proprietary research, unique analytical framing.
If it trusts you but never surfaces your content for the question being asked, the problem is Hooks. It is built in part 4: hooks, across chapter 7 and chapter 8.
A passage of content retrieved and used on its own, separated from the page it came from.
An explicit source attribution in a generated answer: the link or reference an answer engine shows for a claim.
The third ECHO pillar. Structured, extractable, self-contained passages that an AI can lift out and attribute without losing meaning.
What a document adds beyond what already exists on a topic.
The decomposition of one user question into many simultaneous sub-queries, each retrieving different content, which the system then synthesises into a single answer.
Generating an answer from live retrieved documents rather than from training data alone.
Nearby in the glossary
The glossary runs alphabetically. Ground truth comes before Grounding and Hallucination after it.