Is this NFT actually immutable?

What it checks

Can the file change? Read tokenURI of a token that exists and look at the scheme. data: is in the chain. ipfs:// and ar:// are content hashes — the bytes behind them cannot change without changing the hash. https:// is whatever that server returns today.

Can the pointer change? Two independent reads. First the verified source — and, if the address is a proxy, the implementation or facets behind it — grepped for a function that rewrites what tokenURI returns. Then the deployed bytecode, scanned for those functions' four-byte selectors. Proxies are resolved by reading the EIP-1967 storage slot, the address embedded in a minimal-proxy stub, and the diamond loupe, not by trusting a label — and the two kinds are not treated alike. A proxy that keeps its implementation in storage is mutable no matter what its current code says, because the code itself can be replaced. A minimal-proxy clone holds its target as a constant in its own code and can never be repointed, so it is judged on its implementation's setters like any plain contract. Explorers label both “proxy”; most NFT contracts on Base are the second kind.

What counts as a setter? More than setBaseURI. A collection whose art is built entirely on chain still builds tokenURI by calling out to a renderer contract, and a setTraitRenderer or setContractRenderer swaps that contract for a different one — the art is in the chain, the pointer to the code that draws it is not. Three of the first six collections this scanner cleared as immutable turned out to be that case. So the match is on shapes rather than a list of names: any set/update/change of something ending in a renderer, a descriptor, a traits contract, or a *Uri.

Why read the bytecode too? Because “unverified, so nobody can check it” is not true. The selector is in the deployed code whether or not the source was published. It cuts the other way as well: a flattened source file carries functions the deployed contract never exposes, so the two reads are shown separately and you can see where they disagree.

Has it happened? Page the contract's inbound transactions for calls to those functions. Long histories are truncated, and the result says so rather than reporting a clean bill.

Everything runs in your browser against Blockscout and a public Base RPC. No backend, no wallet connection, nothing stored. A setter is not a scandal — reveals and gateway migrations need one. It just means the collection is not immutable, whatever the listing says.