What a token actually proves about ownership

The interesting question about non-fungible tokens is not whether they are valuable. It is what one actually establishes — because that has a clear answer, and it is narrower than almost all of the surrounding discussion assumes.

The mechanism, precisely

A conventional token is fungible: any unit is interchangeable with any other, like a coin. A non-fungible token is one whose units are individually distinguishable, so a contract can track each one separately and record which address controls it.

What the blockchain establishes, with certainty and without trusting anyone: that a specific token identifier within a specific contract is currently assigned to a specific address, and the complete history of that assignment. That is genuine and useful. It is also the entire extent of it.

Crucially, the artwork, video or document is almost never on the chain. Storing large files on-chain is prohibitively expensive, so the token holds a pointer — a piece of metadata containing a link. The token is a ledger entry with an address attached to it.

The three-link chain, and where it breaks

For “I own this thing” to hold, three links must all hold:

Link	Guaranteed by	Strength
Your key controls the token	Cryptography and consensus	Very strong
The token points at the file	A link in metadata	Only as durable as the link
Controlling the token grants rights	A licence or contract, off-chain	Usually absent

The first link is excellent. The second is where the technical fragility lives. The third is where the conceptual problem lives — and it is much the more important of the two.

Link two: pointers rot

If the metadata contains a conventional web address, then the token depends on someone continuing to pay for that server. When it lapses, the token remains perfectly valid and points at nothing. The ledger entry is intact; the reference is dead.

Content-addressed storage improves this substantially: the pointer is a hash of the content, so any copy can be verified as the right file and the reference cannot silently be swapped for a different image. But it still requires that at least one participant is storing the file. Content addressing guarantees integrity, not availability. If nobody keeps a copy, a verifiable pointer to an absent file is what you have.

A small number of projects store the artwork fully on-chain, usually by generating it from code. Those genuinely do not have this problem. They are the exception.

Link three: a chain cannot enforce rights over off-chain things

This is the one that matters, and it is a limit of law rather than of engineering.

A blockchain can enforce anything inside its own state — it can guarantee that only your key moves that token. It has no ability whatsoever to affect the world outside. It cannot stop anyone copying an image, cannot prevent the creator selling the same rights elsewhere, and cannot give you standing to object.

Copyright, in essentially every jurisdiction, sits with the creator unless it is transferred by a written agreement. Buying a physical painting does not give you the right to reproduce it, and buying a token is not different. So what you actually get depends on the licence the seller granted — and the range is enormous:

• No licence at all. Extremely common. You control a ledger entry. You have no rights over the image.
• A personal-use licence. You may display it. You may not make commercial use of it.
• A broad commercial licence. Some projects grant genuinely wide rights, and these are the exception worth noticing.
• Copyright assignment. Rare, and requires a proper written instrument.

The token itself carries none of this. Two visually identical tokens from two projects can convey completely different rights, and nothing on-chain distinguishes them. The only way to know is to read the terms — which many collections never published.

The consequence people find hardest

Because minting is permissionless, anyone can create a token pointing at anyone’s work. There is no gatekeeper, no verification step, and no rights check. So a token pointing at an image is not evidence that the minter had any relationship to it.

Platform verification badges are a social layer bolted on top precisely because the technical layer cannot answer the question. When people say a token “proves provenance”, what it proves is the provenance of the token — the chain of custody of the ledger entry — which is a different and much smaller claim than the provenance of the work.

What this technology is actually good for

The narrowness is not a criticism. A verifiable, transferable, uniquely identified ledger entry that no issuer can revoke is a genuinely useful primitive, and it works best where the entry is the thing rather than pointing at something else.

Fully on-chain generative artwork qualifies: the token and the work are the same object. So does anything where the right lives inside the same system — access to an on-chain application, a position in an on-chain protocol, an identifier used by contracts that can read it. In those cases all three links collapse into the first one, which is the strong one.

Where the token points outward at a physical object, a legal right or a service, the chain is doing a small part of the job and documents are doing the rest. That can work perfectly well. It just is not the chain that makes it work.

Two questions before you buy anything

Where is the content stored, and does it survive the project’s own website going away? And what licence am I granted in writing — with the answer being a document you have read, not an assumption from how the collection is described.

If the answer to the second is “none”, that is a legitimate thing to buy knowingly. It is a bad thing to discover afterwards. Nothing here is advice about whether to buy any of it; see our disclaimer.