JSON Vs JSONB Storage

Summary: json stores an exact text copy of the input; jsonb stores a parsed binary structure — that one difference explains their speed, indexing, and operator support trade-offs.

At a glance

Question	Answer
What is it?	Two PostgreSQL types for storing JSON data
Why does it matter?	They differ in storage format, which changes speed, indexing, and which operators work
When do I use it?	Almost always jsonb, unless exact input formatting must be preserved

How it works

	json	jsonb
Storage	Exact text copy of input	Parsed, decomposed binary format
Whitespace/key order	Preserved exactly	Not preserved
Duplicate keys	All stored (last wins on read)	Only last duplicate kept
Write speed	Faster (no parsing)	Slightly slower (parses on insert)
Read/query speed	Slower (reparses every query)	Faster (already parsed)
GIN indexing	Not supported	Supported (@>, ?, etc.)
Operators (||, jsonb_set, etc.)	Not supported	Supported

Example

-- clarify_history column is jsonb, which is why || and jsonb_set work on it
alter table patch_sessions add column clarify_history jsonb;

If that column were json instead, the \|\| concatenation would not work at all — json has no \|\| operator.

Remember

• Default to jsonb for nearly all use cases.
• Use json only when exact byte-for-byte input (whitespace, key order, duplicate keys) must be preserved — rare, mostly audit-logging scenarios.
• Operators like \|\| and functions like jsonb_set only exist for jsonb.

Related notes

• PostgreSQL JSONB Merge Tools - operators that only work because the column is jsonb