OP_RETURN

OP_RETURN is a Bitcoin Script opcode that makes its output deliberately unspendable. The "spending" path is a guaranteed-fail, so an output starting with OP_RETURN never enters the UTXO set - giving you a way to embed arbitrary data into a Bitcoin transaction without bloating the unspent-output database.

The basic structure: OP_RETURN <data>. The output usually carries zero satoshis, since nothing can ever spend it. The data sits in the chain history forever but doesn't burden the UTXO set.

How much data nodes will relay is policy, not a consensus rule. Bitcoin Core started relaying OP_RETURN outputs with up to 40 bytes of data in version 0.9 (2014) and raised that to 80 bytes in 0.11 (2015). The 80-byte limit stood for a decade. Version 30.0 (October 2025) raised the default to 100,000 bytes, which leaves the transaction size limit as the real cap, and started relaying transactions with more than one OP_RETURN output. Bitcoin Knots keeps the 80-byte default and, by default, applies it to other ways of embedding data as well. Miners have always been free to put larger OP_RETURN outputs in their own blocks.

What OP_RETURN is used for:

• Proof-of-existence / timestamping. Hash a document, embed the hash, prove later that the document existed before that block height. Used by OpenTimestamps and similar services.
• Token protocols. The Omni Layer, where Tether first issued USDT, and Runes carry their token messages in OP_RETURN outputs. Counterparty uses them for smaller messages.
• Commitments from other chains. VeriBlock's miners published that chain's state in Bitcoin OP_RETURN outputs to borrow Bitcoin's proof of work.
• Layer-2 anchoring. Some second-layer protocols anchor state commitments via OP_RETURN.
• Inscriptions (sort of). Inscriptions use Taproot witness data rather than OP_RETURN, so the 80-byte limit never applied to them, but the data-embedding question is the same and OP_RETURN gets caught up in the debate.

The community debate around OP_RETURN:

• Pro-data view: the chain is for whatever its users want to pay fees for. Embedding data has legitimate uses (timestamping, sidechain commitments, etc.) and the fee market handles abuse.
• Anti-data view: Bitcoin's primary purpose is monetary; non-monetary data uses inflate fees for everyone, push out legitimate financial transactions, and shouldn't be relayed or mined. Some nodes (Bitcoin Knots) deliberately filter such transactions from their mempools.

Both positions are defensible; the live argument in 2024-2026 is which node relay policy is appropriate. The protocol itself accepts OP_RETURN; node-level policy is where the contest happens.

See OP_RETURN-based Tokens for the early token-experiment use case. For the full history, from data hidden in fake addresses to Core 30.0, with the RPC calls to find OP_RETURN outputs in a block yourself, read ChainQuery's OP_RETURN: Bitcoin's Data Layer.