Grounding is the practice of tying an AI's output to real, verifiable sources. A grounded answer is built from documents the engine fetched and can point to; an ungrounded one comes from the model's parametric memory, which may be outdated or wrong.
It's the main defence against hallucination, and it's foundational to AI search surfaces like Google's AI Overviews and Perplexity. They retrieve, then answer from what they retrieved, then cite. For users it means more trustworthy, current answers with sources to check.
For brands, grounding is the opportunity: if the engines prefer to answer from sources they can verify, then being a clear, authoritative, reachable source about your own brand and category is how you get pulled into, and credited in. Those answers. Make your facts easy to ground, and you're more likely to be the evidence.