diff --git a/site/docs/recipes/json-traversal.mdx b/site/docs/recipes/json-traversal.mdx
new file mode 100644
index 000000000..5d2929389
--- /dev/null
+++ b/site/docs/recipes/json-traversal.mdx
@@ -0,0 +1,38 @@
+import Tabs from '@theme/Tabs'
+import TabItem from '@theme/TabItem'
+import { Types } from './json-traversal/Types'
+
+# JSON Traversal
+
+In the era of data-driven applications, SQL databases such as PostgreSQL, MySQL,
+and SQLite have evolved to support JSON traversal. JSON provides a flexible and
+dynamic data structure, and these databases now offer native support for storing,
+querying, and manipulating JSON data within the SQL framework.
+
+PostgreSQL, MySQL, and SQLite have introduced native JSON data types, along with
+functions and operators for extracting values, filtering conditions, aggregating
+arrays, and indexing JSON data. This enables developers to leverage the strengths
+of SQL and JSON together, combining relational querying with the flexibility of
+JSON manipulation.
+
+These capabilities have always fascinated us, and finally we are able to bring
+them to Kysely and through some exploration and experimentation, we'll hopefully
+land together on a solution that is both elegant, practical, and most importantly,
+type-safe.
+
+Each dialect Kysely supports at its core, has a different way of handling JSON
+traversal. So this recipe will require choosing a dialect to work with:
+
+<Tabs queryString="dialect">
+  <TabItem value="postgresql" label="PostgreSQL" default></TabItem>
+  <TabItem value="mysql" label="MySQL"></TabItem>
+  <TabItem value="sqlite" label="SQLite"></TabItem>
+</Tabs>
+
+## Defining Types
+
+<Types />
+
+## Querying
+
+## Inference
diff --git a/site/docs/recipes/json-traversal/Types.tsx b/site/docs/recipes/json-traversal/Types.tsx
new file mode 100644
index 000000000..c035dd4f9
--- /dev/null
+++ b/site/docs/recipes/json-traversal/Types.tsx
@@ -0,0 +1,51 @@
+import React from 'react'
+import { Dialect } from '@site/docs/getting-started/shared'
+import CodeBlock from '@theme/CodeBlock'
+
+export interface TypesProps {
+  dialect: Dialect
+}
+
+export function Types(props: TypesProps) {
+  return (
+    <>
+      <p>
+        When defining a JSON column's type, it must follow the following rules:
+        <br />
+        <br />
+        <strong>Root column type</strong> - The root select type must be of
+        object or array type. It can be nullable, but cannot be optional (
+        <code>jsonColumn?:</code>), just like any other column type definition.
+        Its insert and update types must be strings, as you'd{' '}
+        <code>JSON.stringify</code> JSON parameters.
+        <br />
+        <br />
+        <strong>Nested field type</strong> - Nested fields must have a JSON
+        native type (string, number, boolean, null, object or array). Unlike the
+        root column, nested fields can be optional (<code>field?:</code>).
+        <br />
+        <br />
+        <strong>
+          Unknowns, JSON (Discriminated) Unions and other complexities
+        </strong>{' '}
+        - Supporting traversal to not-well-defined JSONs or complex types was
+        not part of phase 1. It could work right now, but we haven't tested it.
+        We'd appreciate any feedback or real-world examples, as we prepare for
+        the next phases.
+        <br />
+        <hr />
+      </p>
+      <CodeBlock language="ts" title="src/types.ts">
+        {`import { ColumnType, Generated, JSONColumnType } from 'kysely'
+        
+export interface Database {
+  person_metadata: PersonMetadataTable
+}
+
+export interface PersonMetadataTable {
+  profile: JSONColumnType<{ created_at: string }> // short for ColumnType<T, string, string>
+}`}
+      </CodeBlock>
+    </>
+  )
+}