Scan reveals 96,000 fake-address outputs clogging Bitcoin’s unspent database

Blockchain data company Bitquery published a census on Sept. 2 of four ways text appears in Bitcoin, scanning 965,135 blocks and revealing that the same words can impose very different costs on nodes.

Its study counted 13,062 OP_RETURN taunt transactions that added no spendable state, against 96,231 fake-address text outputs that remain in the unspent transaction output set.

The March 21-26, 2026, campaign named Bitcoin Core developer Luke Dashjr in the same two-sentence taunt. Bitquery counted 13,062 transactions from 13,062 sending addresses across 579 blocks, with 0.1014 BTC paid in aggregate fees. The chain records the text, timing, inputs, and fees, but it does not identify the payer, author, or motive.

Across the broader scan, Bitquery classified 4,412,782 OP_RETURN outputs as readable text. Their node cost differs from fake-address text because Bitcoin treats the two output types differently after confirmation.

Nodes do not need to retain an OP_RETURN output in the UTXO database used to check whether coins can be spent, although its data remains in Bitcoin’s block history. The Bitcoin developer guide describes this null-data construction as preferable to methods that create outputs which look spendable and burden the UTXO set.

Fake-address text places readable bytes where an address hash would normally go. A node sees an apparently spendable output, while the embedded text has no known corresponding private key.

Bitquery classified 96,231 such outputs across 3,286 transactions and estimated that they hold about 3.2 BTC that is effectively unspendable. Those outputs persist as current node state unless they are spent.

Bitcoin Core 30.0, released in October 2025, made larger OP_RETURN transactions easier to relay under its default policy. The release notes raised the default -datacarriersize setting to 100,000 and allowed multiple OP_RETURN outputs under an aggregate limit for relay and mining.

These standardness choices did not change the consensus rules that determine whether a block is valid.

Bitquery found that messages longer than the former 80-byte payload ceiling appeared only a few dozen times in most months before v30 and rose into the thousands by February 2026. Easier default propagation offers one mechanism for that rise, but the observational sequence does not show that v30 alone caused it.

The March taunt disclosed text is 62 ASCII bytes, which already fit below the former ceiling. The campaign measures the low cost of repeating OP_RETURN text after v30.

Bitquery also lacked coinbase fields for 191,427 historical blocks and did not count witness inscriptions. The study therefore compares the state consequences of four measured channels rather than tallying every piece of text embedded in Bitcoin.

Within that scope, OP_RETURN expands block history without bloating the UTXO set, while fake-address text creates the lasting node-state burden.