JSON Schema is a vocabulary for annotating and validating JSON documents, specified through a series of Internet-Drafts published by the IETF. The specification has evolved through several drafts — Draft-04 (2013), Draft-06 (2017), Draft-07 (2018), Draft 2019-09, and Draft 2020-12 — each adding capabilities while refining the validation model. Unlike JSON itself (which has a single-page grammar), JSON Schema is a substantial specification that defines a rich type system, validation keywords, conditional schemas, and composition mechanisms.

At its core, JSON Schema describes the shape of JSON data using JSON itself — schemas are JSON documents that define constraints other JSON documents must satisfy. The fundamental keyword is "type," which restricts values to one of JSON's types (string, number, integer, boolean, null, object, array). Each type has associated validation keywords: strings have "minLength," "maxLength," and "pattern" (regex validation); numbers have "minimum," "maximum," and "multipleOf"; arrays have "items," "minItems," "maxItems," and "uniqueItems"; objects have "properties," "required," "additionalProperties," and "patternProperties."

The "format" keyword provides semantic validation beyond structural types. A string that must contain a valid email address is described as {"type": "string", "format": "email"}. Standard formats include "date-time" (ISO 8601), "date," "time," "email," "uri," "uuid," "ipv4," "ipv6," and "hostname." Format validation is optional by specification — validators may choose to validate format values or treat them as annotation-only — which creates inconsistency across implementations.

JSON Schema's composition keywords enable powerful schema combinations. "allOf" requires a value to satisfy all listed schemas (intersection), "anyOf" requires at least one (union), "oneOf" requires exactly one (exclusive union), and "not" inverts a schema. These keywords, combined with "$ref" for referencing other schemas, enable building complex type hierarchies from simple building blocks. The "$ref" mechanism also enables recursive schemas — a schema can reference itself to describe tree structures like nested comments or organizational hierarchies.

In the API ecosystem, JSON Schema plays a central role in the OpenAPI Specification (formerly Swagger), which uses an extended subset of JSON Schema to describe API request and response bodies. When you define a POST endpoint's request body in an OpenAPI document, the schema property uses JSON Schema vocabulary. This integration means that a single JSON Schema definition can drive API documentation generation, request validation, client SDK generation, and mock data creation.

JSON Schema validation powers critical infrastructure across the software industry. Configuration file validators use schemas to catch errors before deployment — VS Code validates settings.json against a schema, Kubernetes validates resource manifests, and CI/CD pipelines validate workflow definitions. Schema registries in event streaming platforms like Apache Kafka enforce data contracts between producers and consumers, preventing schema evolution from breaking downstream systems. The ability to automatically generate schemas from sample data — inferring types, detecting formats, and identifying required fields — accelerates the creation of these validation rules, transforming examples into enforceable contracts.