adult-xxx.org arrives in this brief with current_intent set to informational and with primary_asset type custom and an empty ref. That is the only concrete fact the grounding supplies: there are no real_videos rows, no real_creators entries (and therefore no aka names to weave), creator_stats is null so catalogue size cannot be cited, photo_count is not present, categories is an empty list, tags is an empty list, and target_keywords is empty so none can be worked in without invention.
Because the instruction is to use only grounding data for factual claims, this page cannot name a CATEGORY the content sits in, cannot lean on top tags, and cannot open on a filmed moment or a gallery count. The concept is a domain-shaped seed, not a populated creator or title cluster. Sibling pages that inherit actual titles would outrank a stub that pretends those titles exist. The honest job of the copy is to classify the query, state the asset gap, and stop short of fabricating catalogue substance.
Competition for a string that looks like an adult destination domain is typically high even when the local asset graph is empty; conversion weight stays low until a real video, creator, or category is linked. The niche cluster is therefore the query itself—adult_xxx_domain—rather than a thematic shelf drawn from tags. Asset resonance is sparse_custom_stub: a custom type with no ref and no supporting media.
If later grounding attaches titles, descriptions, tags, or a creator_stats block, the body should be rewritten to open on those facts and to name the categories verbatim. Until then, ranking language should stay limited to the seed_query adult-xxx.org and to the documented emptiness of the catalogue arrays. That constraint is the page’s only defensible unique signal and the only way the copy remains truthful under the use-only-grounding rule.
