An untested backup is not a backup. It is a belief about a piece of metal in a drawer.
This is the step almost everybody skips, and it is the one that decides whether all the other steps mattered. People do not usually lose crypto because their seed phrase was cracked. They lose it because word nine was written as “recieve”, or two words were swapped, or the backup was for a wallet they later replaced — and they found out years later, at the exact moment they could no longer fix it.
What actually goes wrong
Real backup failures cluster into a handful of causes, and every one of them is silent until you need the backup:
- Transcription errors. A misspelling, a word that is not in the BIP-39 list at all, or an ambiguous letter that reads as something else years later. l and 1, 0 and O, u and v in hurried capitals.
- Order errors. The words are correct; the sequence is not. Unnumbered backups are especially prone to this once a card gets shuffled or a tile system is dropped.
- Wrong wallet. You backed up the phrase from a wallet you later abandoned, and the funds are in a different one.
- Missing passphrase. The words are perfect, but the wallet also uses a passphrase that was never recorded. The words alone open an empty wallet — which looks exactly like a failed restore.
- Incomplete capture. A 24-word phrase written on a card that only had room for 20.
Note that the last two produce the same symptom: a technically valid restore showing zero balance. Panic follows, and panic is when people start typing their phrase into whatever website promises to help.
The rule that governs every method below
You never test a backup by typing it into a website. Not a “checker”, not a “balance viewer”, not a tool that promises the words never leave your browser. There is no legitimate site that needs your phrase, and the habit of entering it somewhere is precisely what every phishing operation depends on.
Every method here keeps the phrase inside a device you already trust with your keys.
Method 1: the wipe-and-restore (the real test)
This is the only test that proves the whole chain end to end, because it is exactly what you will do in a real recovery.
- Confirm your backup is physically in front of you and legible.
- Reset the hardware wallet to factory settings.
- Restore it using only the written backup — not memory, not the original card if you are testing a copy.
- Check that the first receiving address matches what it was before, and that balances appear.
Step 3 is the whole point. If you find yourself filling a gap from memory, the test has failed and you have just proved the backup is incomplete.
The obvious risk: between the wipe and a successful restore, that backup is the only thing standing between you and losing everything. Do it when you are unhurried, sober and alone, with the device’s original packaging and any second backup within reach. If you hold a meaningful amount, move most of it out first, test, then move it back.
Method 2: the spare-device restore (safer, nearly as good)
If you have a second hardware wallet, restore the backup onto that one and compare the first receiving address against your live wallet. Identical address means the backup reproduces the same keys.
This proves everything the wipe-and-restore proves, without ever putting your working device into a blank state. Wipe the spare afterwards. It is the best option available and the reason a cheap second device is a genuinely good purchase — a point we come back to in hardware wallets.
Method 3: the read-back check (weak, but do it today)
If a full restore is not happening this month, at least do this now:
- Read your backup aloud, word by word, in order.
- Confirm every word is in the BIP-39 wordlist — check against the list itself, from the specification, not by typing it anywhere.
- Confirm the count is exactly 12, 15, 18, 21 or 24. Any other number means something is missing.
- Confirm the numbering is present and unambiguous.
- Confirm your handwriting is legible to someone else. Have a person who does not know the words read them back to you.
That last one catches more errors than people expect. You can read your own bad handwriting; your heirs cannot.
This method cannot detect a wrong order or a wrong-wallet backup, which is why it is a stopgap rather than a substitute.
The checksum will catch some of this for you
BIP-39 builds a checksum into the phrase — four bits for a 12-word phrase, eight for 24. In practice this means a wallet will simply reject most corrupted phrases rather than silently opening the wrong wallet: only about one in sixteen random 12-word combinations passes the check.
That is a useful safety net, with a sharp limitation. The checksum tells you a phrase is well-formed. It does not tell you it is yours. A valid phrase that is not your phrase restores cleanly to an empty wallet, and looks identical to a missing-passphrase failure.
How often
A full restore test when you first set the wallet up, and again whenever something changes: a new backup copy, a new storage location, a passphrase added, a device replaced. Beyond that, an annual read-back check is enough for most people — paired with actually looking at the backup to confirm it is still where you think it is, still legible, and still dry.
Put it on the same calendar reminder as smoke-alarm batteries. It is the same category of task: five boring minutes that only matter on the day everything else has gone wrong.
Sources
- BIP-39 — the wordlist, valid phrase lengths, and the checksum that rejects malformed phrases
- The BIP-39 English wordlist — the authoritative 2,048 words to check yours against