Password Strength Checker
Check how strong a password is and estimate how long it would take to crack, analysed entirely in your browser with nothing transmitted.
Assumes a fast attacker guessing 10 billion passwords per second.
Length beats complexity, and it is not close
A random password's entropy is its length multiplied by the log of its alphabet size. Both terms matter, but not equally — length multiplies, while the alphabet only ever sits inside a logarithm. That asymmetry is why the standard advice is backwards.
| Password | Entropy | Gained over 8 lowercase |
|---|---|---|
| 8 lowercase letters | 37.6 bits | — |
| 8 characters, every class | 52.6 bits | +15.0 |
| 12 lowercase letters | 56.4 bits | +18.8 |
| 16 lowercase letters | 75.2 bits | +37.6 |
Adding upper case, digits and symbols to an eight-character password buys 15.0 bits. Adding four more lowercase letters buys 18.8. The entire "must contain a symbol" rule is worth about 3.2 extra lowercase characters — and unlike the symbols, those characters cost nothing in memorability or typing errors.
The reason is arithmetic rather than opinion. Doubling the length doubles the entropy; doubling the alphabet adds one bit per character and nothing more. Composition rules push on the term that barely moves.
Which is why a rule-counting meter can be out by a factor of a quadrillion
A meter that checks which character classes are present has no choice but to assume the password is random over those classes. Real attackers are not random. They try the common things first: a dictionary word, then obvious letter-for-symbol swaps, then capitalisation patterns, then a short suffix.
| Password | Class-counting meter | Guess-common-first | Overstated by |
|---|---|---|---|
P@ssw0rd1! | 65.7 bits | 15.0 bits | 51 bits |
Summer2024! | 72.3 bits | 13.6 bits | 59 bits |
Tr0ub4dor&3 | 72.3 bits | 30.3 bits | 42 bits |
Being straight about what that second column is: it is a model, not a measurement. It
assumes a particular dictionary position and a particular set of transforms — for
P@ssw0rd1!, the single most common base word, two obvious substitutions, a short suffix. Choose different assumptions and
you get a different number. What does not change is the direction and the scale: any attacker
who tries likely things first arrives at a number vastly below what the meter reports.
The cleanest demonstration that a class-counting meter cannot be measuring strength: it scores
P@ssw0rd1! and a random string of the same shape
identically. It never looks at the letters, so it cannot tell the most common
password in the world from one nobody has ever typed.
Meanwhile correct horse battery staple — four words from a
7,776-word list — is 51.7 bits,
stronger than all three examples above, and would be rejected by most sign-up forms for having
no capital, no digit and no symbol.
And the site's password storage matters more than your password
Every "time to crack" figure — including the one above — is a guess rate multiplied out. The guess rate is not a property of your password. It is a property of how the site stored it, and it varies by a factor of five million between the two ends of common practice.
| Password | Entropy | Fast hash (MD5, SHA-1) on rented GPUs | bcrypt at cost 12 |
|---|---|---|---|
| 8 lowercase letters | 37.6 | 1 seconds | 60 days |
| 8 characters, every class | 52.6 | 9.2 hours | 5 thousand years |
| 12 lowercase letters | 56.4 | 6 days | 76 thousand years |
| 3-word passphrase | 38.8 | 2 seconds | 136 days |
| 4-word passphrase | 51.7 | 5.1 hours | 3 thousand years |
| 5-word passphrase | 64.6 | 5 years | 23 million years |
| 6-word passphrase | 77.5 | 35 thousand years | 1.8e+2 billion years |
A four-word passphrase falls in 5.1 hours if the site stored it as a fast hash, and holds for 3 thousand years if it used bcrypt. Same password, same attacker, six orders of magnitude apart.
Two things follow. If you are choosing a password: four words is fine for a site that stores it properly and thin for one that does not, and you cannot tell which from the outside — so use five or six for anything that matters, and a password manager with long random strings for everything else. If you are storing passwords: your choice of hash moves the number further than any composition rule you can impose on your users.
How to use
- Type or paste a password.
- Read the entropy estimate and the crack-time figure.
- Note which patterns weakened the score.
- Never test a password you actually use on a site you do not trust.
Frequently asked questions
Is it safe to type a real password here?
Here, yes — the analysis runs entirely in your browser and nothing is transmitted. But it is a poor habit generally: most strength checkers online do send the password, and you cannot tell by looking. Test a similar password rather than a real one anywhere you have not verified.
How is crack time estimated?
From the size of the search space and an assumed guessing rate. Both assumptions matter enormously — an offline attack on a fast hash manages billions of guesses per second, while an online attack against a rate-limited login manages a handful. The same password can be trivial or impossible depending on which.
Why does my complex password score badly?
Because complexity theatre does not fool a modern cracker. Substituting 3 for e, appending a digit and a symbol, or capitalising the first letter are all patterns encoded in cracking rules, so they add far less than the character count suggests. Length and unpredictability are what count.
What do the strength meters on websites actually measure?
Often only whether composition rules are satisfied, which is close to meaningless. A password meeting every rule can be among the most common in existence. Meters based on real pattern analysis, such as the zxcvbn approach, are considerably more honest and will happily rate a long random phrase above a short cryptic one.
Does it matter if my password is in a breach list?
Enormously, and it overrides every other consideration. A password that has appeared in any breach is tried first in credential-stuffing attacks regardless of how complex it looks. Checking against a breach corpus is more useful than any strength score.
What should I do with a weak result?
Replace the password rather than patching it, and let a password manager generate and store the new one. Extending a weak password keeps the predictable core that made it weak, which is why appending characters is a poorer fix than starting again.
🔒 This tool runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.