Web Grounding
When an AI engine bases its answer on a live web search rather than on its trained-in memory alone.
What is web grounding?
Web grounding is when an AI engine bases its answer on a live web search rather than on its parametric memory alone — retrieving current sources and citing them, instead of relying only on what it learned in training.
Web grounding is the moment an engine stops reciting from memory and starts reading the live web — the moment your content can win a citation.
A grounded answer runs a retrieval step: the engine searches the web, pulls relevant pages and composes its reply from them, usually with links back to the sources. This keeps answers current and lets the engine cite real URLs, which is exactly where a brand's content can be selected and quoted.
Grounding behaviour varies by engine and query. Perplexity is always grounded; ChatGPT, Claude and Gemini ground some queries and answer others from parametric memory. Because the grounded path rewards retrievable, well-structured content, it is the surface AEO and GEO most directly influence.
SkuLift measures grounded and parametric visibility separately, so a brand can see where it wins through retrieval versus where it is embedded in the model — and invest its content effort accordingly.