What does fully on-chain NFT mean?
A precise way to distinguish on-chain ownership, metadata, media, generation, and the external dependencies that an NFT may still require.
“On-chain” can describe different layers
Every conventional NFT records ownership and transfer rules on a blockchain. That alone does not mean the token’s name, image, or animation is stored there. A useful inspection separates at least four layers: ownership, metadata, media, and contract control.
On-chain metadata
Metadata is on-chain when the contract returns the document from chain state or derives it through contract execution, commonly as an inline JSON data: URI. If the returned JSON points to off-chain media, only the metadata layer is on-chain.
On-chain media
Media is on-chain when the bytes or deterministic instructions needed to render it are available through contract state and execution. Inline SVG is a common example. Some generative works store traits or seeds on-chain but rely on an external renderer; that is a meaningful design, but the renderer remains a dependency unless its required code and assets are also durable.
Dependencies can cross contracts and chains
A token may call another contract, resolve an ENS name, request CCIP-Read data, or point to a web3:// resource. That dependency has its own owner, proxy, verification status, and chain assumptions. The original collection can be immutable while the target remains changeable.
Chain permanence is also not absolute. Access depends on nodes, clients, indexing software, and—on some networks—sequencers or bridges. NFT Inspector records these dependencies separately instead of treating “on-chain” as a binary badge.
A practical definition
A strongly fully on-chain NFT can be reconstructed from the relevant chain state without a project-operated server, mutable name, or unavailable private code. The contract paths that produce it should also be understood: an upgradeable or privileged contract can keep all bytes on-chain while retaining the ability to change them.
This is why storage permanence and trustlessness are reported as separate dimensions. Location answers where the data lives. Control answers who can change the system that returns it.