Storing a Seed Phrase: Paper, Steel and What Fails

Most people treat seed phrase storage as a stationery problem: buy a pen, write twelve words on a card, put the card in a drawer. That framing is why backups fail. Seed phrase storage is a threat-modelling problem, and the threats are dull ones — fire, damp, sunlight, a house move, a relative who tidies up.

The phrase is not a password reminder. It is the wallet. Anyone who reads it can rebuild the keys and move everything, with no recovery email and no support line. Every backup decision follows from that: you are protecting a bearer asset, not a login.

Key takeaways

• Paper is easy and fragile. It burns, it soaks, it fades in sunlight, and it is legible to anyone who finds it.
• Steel survives fire and water but costs money and is easier to stamp wrong than to write wrong.
• Splitting a phrase is a plan, not a trick. Halves in two places need a documented reassembly path, and each half alone is still a liability.
• Digital copies are the most common failure. A photo or a synced note turns the phrase into a secret stored on a device that can be compromised remotely.

What seed phrase storage is really protecting

It is protecting a bearer key, not a document. A bank statement has a name on it and a phone number to call. A seed phrase has neither. Ownership is whatever the phrase says, which is why the same twelve words that rescue you from a broken phone also hand your balance to a stranger who reads them off a card.

That creates two failure directions, and good storage holds both. You fail if the backup is destroyed — fire, flood, a discarded box during a move. You also fail if it is found — by a burglar, a housemate, a cleaner, or a backup service that indexes it.

These pull in opposite directions. Every copy you add reduces destruction risk and raises discovery risk, so no setup minimises both. The realistic goal is to make destruction unlikely and discovery useless — and to be honest about which of the two your current arrangement favours.

Paper: cheap, legible, and short-lived

Paper is the default because it costs nothing and needs no tools. It is also the weakest link in most setups, and the failures are predictable enough to list.

What paper survives

Paper survives obscurity. A card taped inside the back cover of a book, or folded into a folder of tax documents, is not where a smash-and-grab looks first. Cheap paper also survives being replaced — you can recopy a faded card in minutes, which makes paper the easiest format to refresh.

It is also easy to check. You can read it in ten seconds, compare it against your device, and put it back. For a balance you would shrug off losing, a paper copy in an unremarkable place is reasonable — as long as you treat it as a copy, not a guarantee.

What paper does not survive

Paper loses to water, fire, and ultraviolet light, and it loses to all three without announcing itself. Thermal paper fades to blank in months. Inkjet ink runs at the first damp. Ordinary notebook paper in a humid drawer can grow mould that eats the ink from behind. A card readable in 2021 may be a grey smear when you need it in 2026, and you will not know until you check.

The other failure is disclosure. Paper has no access control — whoever picks it up is now a keyholder, whether or not they know it. Labelling the envelope “crypto seed” turns an ambiguous card into an instruction.

If you have not read your backup aloud against your device in the last year, you do not know whether it still works. You know that it once did.

Steel and splitting: paying for durability

Steel plates exist because paper loses to the two disasters people actually fear: fire and flood. They are the right answer for a meaningful balance, and they are not the frictionless upgrade the marketing suggests.

What steel buys

Stamped or engraved metal survives a house fire that destroys everything around it, and it survives being submerged. It does not fade or grow mould, and it does not care about sunlight. Most plates are also small enough to store where paper would be awkward, which quietly improves the obscurity half of the problem.

What it buys is durability, not correctness. A plate with one letter wrong is worth less than a smudged card, because it looks authoritative.

What steel still costs you

It costs money, tools, and attention. Stamping letter by letter is slow, and the failure mode is a mark that is almost right — a word that is valid in the word list but not the word you meant. Every character has to be verified against the device, then re-verified by someone who is not tired. Thin plates distort under the punch, and cheap letter sets leave shallow marks that blur. Treat the first plate as practice, and test-restore from the finished one.

Splitting and reassembly

Spreading a phrase across two locations is the standard answer to the fire problem, and it fails more often than people expect — not because the halves are lost, but because nobody wrote down how they go back together. Half a phrase in a drawer and half with a relative is a plan only if both halves can be found, identified, and combined by someone who is not you, on a bad day.

Each half alone is also a liability of a different kind. A single missing word shrinks the guesswork enormously compared with a full phrase, and the person holding half a phrase may not treat it with the care the maths deserves. If you split, document the split somewhere that is not itself the phrase, and have the person who would actually have to use it rehearse the reassembly.

Paper vs steel vs split vs screen: a seed phrase storage comparison

The formats differ less in how well they hold ink than in which failure they trade away.

Choosing a seed phrase storage setup that survives you

Start by counting exposure rather than copies. Most people believe they have “three backups” when they have three complete keys, on more systems than they realise, all in one building.

1. Count your copies. The card in the drawer, the photo in your camera roll, and the note in your password manager make three — not one backup in three places, but three complete keys.
2. Count the systems each copy touches. The drawer is one. The phone is one, plus whatever it syncs to. The password manager is one, plus its own device backups. Three copies routinely sit on five or six systems.
3. Multiply. Three copies on six systems is up to eighteen separate chances for one readable copy to be seen, copied, indexed, or photographed by someone else. The word “redundant” hides that number.
4. Subtract the copies that survive a fire. If all three live in the same building, that answer is zero, and the eighteen exposures bought you nothing at all.
5. Write one page that says where each copy is, what it is missing, and how it is put back together. Store that page somewhere separate from every copy, and give it only to someone who needs it.

Then test it. Restore the wallet on a spare device, or in an app that supports watch-only import, before you put value behind it. Test-restore is the only step that converts an assumption into a fact, and it catches the failure that matters most: a phrase that is complete, legible, and wrong.

If your wallet supports a passphrase, decide deliberately whether you are using one. It adds a factor that is not written on the plate, which is genuinely useful, and a second thing that must never be lost — a correct phrase with the wrong passphrase opens an empty wallet. Record separately that a passphrase exists, so nobody concludes the phrase is broken and gives up.

Decide, too, what happens when the device is gone. A backup only the hospitalised or locked-out owner can interpret is not a backup, it is a delay. One person, one sealed instruction, one documented location is usually enough.