What is a grounding rate for AI content?
Grounding rate is the share of AI-generated work in which every claim, figure and quote traces back to an approved source, measured before the work ships.
Also called: grounding score.
Why it matters.
A brand score tells you whether work sounds and looks right; a grounding rate tells you whether it's true. Both matter, and the second is the one that leads to corrections, complaints and legal reviews when it's missed.
Tracking the rate over time shows whether your context is doing its job. A falling rate usually means agents are being asked about things the approved sources don't cover yet, such as a new product or segment, and points to where the claims library needs work.
Example.
How Brail handles it.
Brail checks each piece for claims that don't trace back to your approved sources and flags them before they ship. Over time you can see how often work passes and where missing proof points cause the most flags.
Questions people ask.
- How is grounding rate different from a brand score?
- A brand score measures how well work matches the brand; grounding rate measures whether its claims are backed by approved sources. A piece can score well on one and badly on the other.
- What grounding rate should we aim for?
- For anything published, every claim should be grounded, so the goal is that nothing unsupported ships. The rate before review shows how often agents need correcting, and it should rise as your sources improve. See approved claims library.
Related terms.
- Grounded output: Grounded output is AI-generated work in which every factual claim, figure and quote traces back to an approved source, rather than to whatever the model guessed or remembered.
- Approved claims library: An approved claims library is the set of statistics, customer quotes, results and statements an agent may use, each with its source, its owner and any rules about when and how it applies.
- Brand score: A brand score is a number that shows how closely a piece of output matches a brand, based on an automated check against its rules.
- Quality floor: A quality floor is the minimum standard every piece of output has to meet before it ships, whoever or whatever made it.
- Context coverage: Context coverage is the share of a team's common AI requests that have the context, examples and assets they need in a governed source, so agents don't have to guess or fill gaps.