
The NFT metadata problem doesn't show up at mint time. It shows up months or years later, when a collector opens their wallet and sees a broken image icon where their NFT used to be. The token is still there, the transaction still happened, the ownership record is still perfectly intact on-chain. What's missing is everything that made the NFT actually mean something visually, and understanding exactly why that happens, and why it's so hard to fix after the fact, is the case for taking pinning seriously before it becomes a problem instead of after.
TL;DR:
tokenURI after minting, so there's often no "just fix the link" option once the underlying content goes offlineIt's rarely dramatic. A marketplace listing shows a grey placeholder square instead of the artwork. A wallet app shows a token with no image and a name field reading "Unknown" or nothing at all. Sometimes it's a loading spinner that never resolves, the request for the metadata or image technically went out, but nothing answered it. None of this touches the blockchain layer at all, the ownership record, the transaction history, the token ID, all of it stays exactly as valid as the day it was minted. What's broken is entirely off-chain: the JSON metadata file, the image file, or both, sitting at a CID that used to resolve and now doesn't.
A broken link on a regular website is usually fixable in minutes: update the URL, redeploy, done. An NFT doesn't have that same escape hatch by default. Many of the most common, straightforward ERC-721 contract implementations store the tokenURI at mint time with no built-in function to change it afterward, immutability by design, the same property that makes an NFT's ownership record trustworthy in the first place. That means fixing a broken metadata reference isn't a five-minute deploy, it can require a contract upgrade path that may not exist, or simply may not be possible at all for a standard, non-upgradeable contract. The mint already happened. The reference is already set. If what it points to disappears, there frequently isn't a clean way to point it somewhere new.
This isn't a hypothetical edge case dreamed up to sell pinning services. Content and link rot is a well-studied phenomenon on the web generally: Pew Research Center found that 38% of webpages that existed in 2013 are no longer accessible today, over a third of the open web, gone within about a decade. An NFT's off-chain metadata and image are subject to the exact same decay dynamic as any other piece of hosted content, unless something is actively counteracting it. The only thing that changes for an NFT specifically is that the address pointing at that content, unlike a normal URL, was written permanently into a blockchain transaction rather than a webpage someone could quietly update.
This is the part that gets misunderstood most often: using IPFS for NFT metadata is not, by itself, a persistence guarantee. This series' content-addressing article covers the exact mechanism, a CID is a valid, permanent, mathematically correct address for a specific piece of content, but the address staying valid says nothing about whether any peer on the network is actually still holding that content. An ipfs:// reference in a tokenURI field is real and correctly formed even after every peer that ever hosted the underlying file has gone offline, it just resolves to nothing. Minting on IPFS moves the failure mode; it doesn't remove it.
The specific, common way this actually happens: a project uploads and pins its collection through a free trial, a personal IPFS node running on a laptop, or a pinning subscription that later lapses, and everything works fine at launch, mint day, and for a while after. Then the free trial ends, the laptop gets shut down, or the subscription quietly expires, and nothing dramatic happens immediately, the content just stops being actively pinned anywhere and eventually falls out of the network through ordinary garbage collection. Nobody notices for weeks or months, until a collector happens to load the token and sees a broken image where an actual piece of art used to be.
The blockchain-level facts don't change: the token is still owned, the transaction history is still intact, the smart contract still works exactly as deployed. None of that matters to a collector looking at a broken image in their wallet. A project with widespread broken metadata reads as abandoned or careless regardless of how sound the underlying contract is, and that reputational damage doesn't require a large percentage of a collection to actually break before it registers, a handful of visibly broken tokens circulating on secondary markets is enough to raise doubt about the whole collection's reliability.
Because fixing a broken tokenURI after the fact is often difficult or impossible, the entire problem this article describes only has one reliable answer: pin the asset and metadata with infrastructure that stays online on an ongoing basis, before the mint transaction ever goes out, not as a reaction once something's already broken. The exact mechanics of doing that correctly on AIOZ Pin, including the common failure mode of pinning an asset and its metadata as two separate operations that can succeed or fail independently, are covered in this series' companion articles on why pinning matters and how to mint with persistence built in from step one.
What does a broken NFT actually look like to a collector? A placeholder image, a spinner that never resolves, or a listing with missing attributes, while the token's on-chain ownership record stays completely valid and unaffected.
Can I just fix a broken NFT's metadata link after minting? Often no. Many common ERC-721 contracts have no built-in function to change a tokenURI after mint, so there's frequently no simple way to repoint a broken reference once it's set.
Is content and link rot a real, measured problem, or just a risk specific to crypto? It's measured well beyond NFTs. Pew Research Center found 38% of webpages that existed in 2013 are no longer accessible today, the same underlying decay dynamic unpinned NFT content is subject to.
Does using IPFS for NFT metadata guarantee it stays available? No. A CID stays a valid, correctly-formed address forever, but that says nothing about whether any peer is still actually hosting the content it points to.
Why do NFT collections break months or years after mint instead of immediately? Because the underlying content is usually pinned somewhere that works fine initially, a free trial, a personal node, then quietly stops being maintained, and the content only becomes unreachable once garbage collection eventually reclaims it.
How do I actually avoid this for my own NFT project? Pin the asset and metadata together, with infrastructure built to stay online on an ongoing basis, before minting, not after. This series' minting tutorial covers the exact workflow.

ERC-1155 metadata on IPFS breaks when filenames ignore the {id} rule. Here is the exact 64-character format, a working folder layout, and the per-token option.

AIOZ Pin rate limits in one place: 180 requests per minute for the API, 15 per CID on the public gateway, and no predefined limit on premium gateways.

IPFS pin stuck on Pinning? Here is what Pinning, Pinned, and Failed mean on AIOZ Pin, why big directories can sit there, and how to read a pin's real state.

Two identical files, three IPFS tools, three different CIDs. Here is why that happened for years, and how IPIP-0499 made content addressing deterministic.

AIOZ Pin MCP server setup has two paths: llms.txt lets a super intelligence assistant read the API, the MCP server lets it call 11 tools. Here is each one.

AIOZ Pin bills three flat rates against a token wallet, with no tiers to pick between. Here is what storage, bandwidth and pinned files cost, with examples.