In one sentence
Grounding is tying a language model's answer to external sources, such as live web results or a document set, so that its claims can be checked against something outside the model.
What it means
A model answering from memory can be fluent and wrong. Grounding supplies it with retrieved text and asks it to answer from that text. Retrieval-augmented generation is the usual method.
Grounding reduces errors but does not remove them. The model can misread a source, choose a poor one, or blend a source with its memory. For brands the important point is that grounded answers depend on what is retrievable right now, so your current pages and coverage matter more than they do for answers from training data.
What to do about it
Make sure the pages a grounded answer would need exist and are crawlable: product facts, pricing, comparisons, documentation. Keep them current. Check that AI crawlers are not blocked.
When you test, compare an answer with search enabled against the same question without it. The difference shows how much of what the model says about you depends on retrieval.
How Pineprompt measures it
Pineprompt records the sources returned with each answer, which shows what the answer was grounded in on platforms that expose it. Coverage differs by platform, so see the methodology.
Frequently asked
- What is Grounding?
- Grounding is tying a language model's answer to external sources, such as live web results or a document set, so that its claims can be checked against something outside the model.
- How does Pineprompt measure Grounding?
- Pineprompt records the sources returned with each answer, which shows what the answer was grounded in on platforms that expose it. Coverage differs by platform, so see the methodology.