Category taxonomy uses controlled vocabulary, not freeform text

Free-text categories produce Restaurant, restaurant, Restaraunt, and Food as four distinct buckets for the same concept. Your browse pages fragment, faceted filters break, sitemaps balloon with near-duplicate URLs that Google treats as keyword cannibalization, and analytics reports become impossible to trust. Once inconsistent data lands in the database, deduplication is a manual SQL cleanup job that costs far more than enforcing a controlled vocabulary at submission time. Schema.org Category emitters also require stable enum values.

Replace the free-text input with a select backed by a canonical enum or category table, and validate the submitted value against that list on the server. Store categories as a foreign-key reference or a constrained string column so the database rejects rows with values outside the vocabulary. Define the enum in src/lib/taxonomy/categories.ts and import it in both the form and the API handler.

export const CATEGORIES = ['restaurant', 'cafe', 'retail', 'services'] as const
export const CategorySchema = z.enum(CATEGORIES)

Seed a category table with the same values and add a CHECK constraint or FK.