MongoDB, JSON, and BSON: What Actually Gets Stored

MongoDB stores BSON, not JSON

When you write a document to MongoDB, the driver serializes your JSON-like object into BSON — Binary JSON — for storage and transport. BSON is a binary-encoded representation of JSON data with extensions. The shell and drivers show you JSON-looking output, so it feels like JSON, but the on-disk and on-the-wire format is BSON. Understanding this explains several behaviors that otherwise look surprising.

Why BSON instead of plain JSON?

BSON is built for databases. It prefixes documents and fields with type and length information, so a query engine can skip and traverse fields without parsing the whole document — much faster than scanning text JSON. It also encodes binary data efficiently. The trade-off is human readability: BSON is not meant to be read directly, which is why tools translate it back to JSON for display.

Types BSON adds over JSON

• ObjectId — the 12-byte identifier MongoDB assigns to the _id of every document.
• Date — a real 64-bit datetime type, which JSON lacks entirely.
• Distinct number types — 32-bit int, 64-bit long, double, and high-precision Decimal128, where JSON has only a single number type.
• Binary data (BinData) — for storing images, files, and UUIDs efficiently.
• Other types like Timestamp and Regular Expression that have no JSON equivalent.

The integer-versus-float problem

JSON does not distinguish integers from floating-point numbers — it has one number type. BSON does, which is why a value can round-trip incorrectly if you treat MongoDB data as plain JSON: a 64-bit integer ID might be coerced to a double and lose precision. When exact numeric types matter, rely on BSON's distinct types (and Decimal128 for money) rather than generic JSON numbers.

Extended JSON for import and export

Because JSON cannot represent ObjectId, Date, or Decimal128, MongoDB defines Extended JSON (EJSON): a JSON-compatible way to encode BSON types using special wrapper keys. It comes in two modes — canonical, which is precise and preserves exact types for round-tripping, and relaxed, which is more human-readable but lossy for some types. mongoexport and drivers use Extended JSON so you can move data through plain JSON tooling without losing type information.

{
  "_id":       { "$oid": "65f0c3a1b2c3d4e5f6a7b8c9" },
  "createdAt": { "$date": "2026-05-08T14:30:00Z" },
  "amount":    { "$numberDecimal": "19.99" }
}

Working with MongoDB JSON

When you export documents or paste them from the shell, you get JSON (often Extended JSON) that you can inspect with normal tools. A formatter makes a dense export readable, a validator confirms it is well-formed, and a tree viewer helps you navigate large nested documents. Just remember that the ObjectId and Date wrappers you see are Extended JSON encodings of BSON types, not plain JSON values.