Find out how strong a password really is: a score, its entropy in bits, an estimate of how long an offline attacker would take to guess it, and specific feedback on what is weakening it. The estimate is not just length times character classes — that number is badly misleading on its own, because 'P@ssw0rd1' looks respectable by that measure and is cracked instantly in practice. Common passwords are caught even behind letter-for-digit substitutions and trailing digits, and repeated characters, sequences, keyboard runs and the familiar capital-word-then-digits shape all cost the strength they appeared to add. The single most important thing worth saying up front: this runs entirely in your browser, and the password is never transmitted, saved, or logged. Typing a real password into a website should make you suspicious, so it is worth checking that claim — open your network tab and watch nothing happen.
| Entropy | Rating | Offline crack time | Example shape |
|---|---|---|---|
| Under 28 bits | Very weak | Instantly | password, 123456 |
| 28–35 bits | Weak | Seconds to minutes | Summer2024 |
| 36–59 bits | Fair | Hours to months | 8–10 mixed characters |
| 60–79 bits | Strong | Years | A 14-character mixed password |
| 80+ bits | Very strong | Longer than is worth anyone's while | A 5-word random passphrase |
No. The analysis runs entirely in your browser — there is no request, no logging, and nothing written to storage. You can verify it: open your browser's network tab and type a password; nothing is sent. Being sceptical of any site that asks for a password is the correct instinct.
It is a measure of how many guesses an attacker would need. Each additional bit doubles that number, so 60 bits is a thousand times harder than 50. It is a better guide than a coloured bar, because it tells you how much a change actually bought you.
It assumes an offline attacker with commodity GPUs managing around ten billion guesses per second against a fast hash. That is deliberately pessimistic — a properly salted, slow hash like bcrypt or Argon2 makes the real figure far higher — but it is the assumption worth planning against, since you cannot control how a site stores your password.
Because length only helps when the content is unpredictable. A long password built from a dictionary word, a sequence, or a keyboard run is found by pattern-based cracking long before a brute-force search would reach its nominal entropy. The warnings list tells you exactly which pattern was detected.
Usually, yes — they are far easier to remember and length buys more entropy than symbol substitution does. Four or five genuinely random words beat a short password with punctuation scattered through it, provided the words are actually random rather than a phrase someone might guess.
Use a password manager and let it generate long random passwords you never have to remember or reuse. Reuse is the bigger risk in practice: a strong password used on twenty sites is only as safe as the worst-run of those twenty.
Strength starts as charset entropy — the password's length multiplied by the base-2 logarithm of the search space its character classes imply — and is then reduced for the patterns that make a password guessable long before that figure suggests. A match against the passwords that top every breach corpus caps entropy at eight bits regardless of length, because a wordlist hit is found in seconds; the match is checked after stripping trailing digits and reversing common letter-for-digit substitutions, so 'P@ssw0rd' and 'password123' are caught too. Runs of repeated characters, ascending or descending sequences, and straight runs along a keyboard row each cost the entropy they falsely contributed, and the very common capital-word-then-digits shape is penalised because cracking tools try it first. Crack time assumes an offline attacker managing ten billion guesses a second against a fast hash, which is deliberately pessimistic — against a properly salted slow hash the real figure is far higher, but the pessimistic number is the one worth planning around. The password is held in component state only: it is never transmitted, never written to storage, and never logged.