A soulbound token is a token that, once it lands in an account, cannot be transferred or sold – it is bound to that account, or "soul." The term was popularised by Vitalik Buterin's January 2022 essay Soulbound, which borrows it from World of Warcraft, where the best items become soulbound the moment a player picks them up. For a DAO the idea cuts to the core of governance design: if the right to govern cannot be bought, then voting power tracks who earned it rather than who can afford it.
Four years of production use have tested that claim, and the largest test – Optimism's Citizens' House – has since rewritten almost every assumption the original design rested on. This page covers why non-transferability matters, the "Decentralized Society" vision that formalised it, the standards and live deployments, what the Optimism record actually shows, the open criticisms, and how Caper uses a soulbound vote token.
Ordinary governance tokens are freely tradable, which quietly turns governance into a market: whoever values control most can simply buy the votes. Buterin's argument is that for governance rights, “transferability is counterproductive” – because concentrated interests are the ones most willing to accumulate governance from everyone else, transferable votes flow toward the actors most likely to abuse them. This is the plutocratic failure mode that token-weighted voting struggles to escape.
Vote-escrow (ve) tokenomics is a partial answer: locking a token for a fixed term makes the resulting vote weight temporarily non-transferable. But the lock can leak back into a market – Aerodrome's veAERO position is itself a transferable NFT, so the "locked" vote acquired a spot price and could be sold outright. Soulbound tokens close that gap by making the credential permanently non-transferable at the token level, not just time-locked.
The concept was formalised in the 2022 paper “Decentralized Society: Finding Web3’s Soul” by Puja Ohlhaver, E. Glen Weyl, and Vitalik Buterin (SSRN 4105763). Its claim is that Web3 so far encodes only transferable, financialised assets, and that a richer society needs non-transferable soulbound tokens (SBTs) to encode the commitments, credentials, and affiliations a "Soul" accumulates over time. Because an SBT cannot be bought or sold, it can carry provenance and reputation that a tradable token cannot.
The paper lists governance applications directly: SBTs enable Sybil-resistant governance (an identity you cannot rent), community-based key recovery, and more honest quadratic funding, where matching weight depends on distinct souls rather than distinct wallets. In each case the point is the same: non-transferability lets a token stand for who you are and what you have done, not merely what you hold.
The minimal on-chain standard is ERC-5192, which extends ERC-721 with a single locked() view and an EIP-165 interface flag so a wallet or marketplace can tell a token is non-transferable. Several DAOs run soulbound credentials in production:
OP gain nothing in this house beyond the single badge, a deliberate check the Token House cannot provide. Its powers now run wider than retroactive public-goods funding: Optimism's own FAQ lists approval of the Collective Intent, veto power over protocol upgrades, and election of representatives to the Developer Advisory Board. How that badge is handed out has changed completely – see the next section.veCRV lock is non-transferable by design – the closest ve-era ancestor of the soulbound idea – even though later forks re-financialised the position (see vote-escrow).Optimism has run the largest and best-documented soulbound governance experiment in production, and it publishes its own post-mortems. The arc is not the one the 2022 papers projected.
Citizenship began as a trust graph. Optimism's history of the House records that Round 1 began with “24 participants selected by the Optimism Foundation,” Round 2 expanded “to 71 participants, with existing Citizens inviting new members,” and Round 3 grew “to 146 participants through continued invitation”. Citizens were chosen, in the Foundation's own framing, "because they were most likely to meet these goals—rather than because they held an intrinsic right to representation."
Round 6 tested the trust graph against a control. The Foundation reports that “no significant differences were found between badgeholders (selected via the web-of-trust model) and a random community sample”, and that both groups "tended to represent an 'enthusiast' persona—those with time and interest in governance who weren't necessarily Superchain stakeholders." On the dimensions Optimism measured, an invitation graph built out of soulbound credentials selected the same kind of person a lottery would have. Round 5 had already found something adjacent: expertise improved answers to "which projects are most useful?" but not to value-laden questions about what the Collective should prioritise.
The House was then rebuilt around eligibility rather than referral. Three classes of Citizen now register directly, and the criteria are published:
Optimism also removed the incentive: “There will be no financial rewards or incentives for being a Citizen or voting in Season 8 and beyond”.
Permanence. “Citizenship eligibility is recalculated at the beginning of every Season, and criteria are subject to change”; the FAQ puts it plainly – “Being a Citizen now doesn't guarantee future Citizens' House membership”. A soul that expires every season is a subscription, not a soul.
Issuer independence. “If you change your governance wallet after receiving the Citizen badge, the badge will be automatically revoked and re-issued to the new wallet”. That one sentence dissolves the key-loss problem the DeSoc paper worried about, and it does so by conceding the point the same paper flagged as the hard one: the issuer can revoke and re-mint at will. Non-transferability binds the holder. It does not bind the issuer.
Non-transferability keeps a credential off the secondary market. It says nothing about the primary one. A chain earns a seat in a one-member-one-vote house by contributing at least 2% of Superchain revenue share, or by ranking in the top fifteen on the same measure. Nobody can buy the badge from a holder; eligibility for it is denominated in revenue.
That is the general lesson for any DAO reaching for soulbound governance, and it is not an argument against it. Making a credential non-transferable genuinely closes the resale market, which is where the crudest vote-buying lives. What it cannot do is remove the contest – it moves the contest to whatever test the issuer uses to decide who qualifies. Design the eligibility test with the same suspicion you would apply to a token distribution, because it is one.
Soulbound governance is not a free lunch, and the DeSoc authors flag most of the tensions themselves. Four years of deployment have moved several of them from theory to record:
Caper uses a soulbound vote token as the on-chain record of participation. A member earns one for each ballot they cast in a caper, and nothing else earns one: trades and transfers never create it. Only the caper itself can deposit the token into an account, so a holder can never pass one to another member, and the token leaves an account only by being burned when its holder leaves.
The record is not the whole of a member's say. Vote weight comes from the tokens a member holds and their vote record together, so a bag bought or transferred in carries no vote weight until its holder votes, and a long voting record with no tokens carries none either. The same weight decides what a member takes when they leave: a member who has voted can exit any time for their fair share of the treasury. The part of governance that has to be earned is transferable to nobody, and the stake that keeps members invested still counts. Voting mechanisms covers the weighting.
The eligibility question the Optimism record raises has a narrow answer here: the only test for earning a vote token is having voted in the caper. There is no committee, no season and no revocation path, and equally no proof of personhood, so a caper's soulbound record resists Sybil attacks only as far as holding its tokens and paying to vote does.
A note on the Optimism citations. The handbook that used to live at community.optimism.io/citizens-house/… has been folded into docs.optimism.io, and every old path in that subtree now returns HTTP 200 while serving the same unrelated protocol-upgrades page. Those URLs look alive to any link checker and are not. This page therefore cites the handbook's source files in Optimism's community-hub repository, which are canonical and resolve correctly. Read 11 August 2026.