Amazon's hidden backend keyword field ('Generic Keywords') indexes a listing for search terms that never appear in the customer-visible title, bullets, or description.
It has two hard limits: 2,500 characters total, and 500 characters per line. Keywords that don't fit naturally into readable visible copy — including the mismatched-intent tier from the Two-Tier Keyword List Strategy — go here instead, per the Two-Step Conversion-Then-Indexing SOP. Keywords already used verbatim in the visible listing don't need to be repeated here; see Keyword Repetition De-Indexing Penalty for the risk of over-repeating terms.
Low-volume, low-relevancy keywords that fall into a Data Dive Four-Bucket Keyword Sorting System "Residue" tier and don't fit naturally into visible title/bullet copy don't need to be forced in — they can be added as individual single words in the backend search-terms field instead, relying on Amazon's ability to match backend terms in combination with on-page terms (see NPO (Noun Phrase Optimization) for how Amazon extends matching beyond a listing's exact visible phrasing).
Because Amazon infers connections between a root word (see Data Dive Roots Feature (Bulk Keyword-Group Exclusion & Prioritization), Cosmo's 15 Semantic Relations) and its long-tail variants, a single lower-volume word (e.g. 'fantasy') can be placed in backend search terms instead of stuffing the full phrase (e.g. 'fantasy booknook') into the title or bullets — Amazon connects it to the ranked root term algorithmically, letting sellers prioritize only the highest-volume root-adjacent words in visible copy.