An informational query seeks an explanation or description of a concept. It decomposes minimally — the model may generate one or two sub-queries to ground the answer, but the fan-out is shallow. These queries are most vulnerable to parametric knowledge dominating because the model often has a high-confidence answer and retrieval is confirmatory rather than generative.
For well-established topics, a model answering an informational query from parametric knowledge will not retrieve anything — which means retrieval optimization does not help. Parametric presence — Wikipedia, widely-cited publications, training data representation — is the lever. For newer topics or repositioned brands where parametric knowledge is absent or wrong, retrieval content provides the correction.