Special characters help a password only when they are the right ones. These five mistakes lock people out or weaken the password.
The short version
Start here if you only have one minute. The full guide below explains each point.
Length beats symbols
A long passphrase outperforms a short decorated password.
Never paste from a document
Invisible characters travel and cannot be retyped.
Styled letters are dangerous
Normalisation can change what actually gets stored.
01
Mistake one: using styled letters
Cursive and bubble letters are separate Unicode characters. Some systems accept them, many reject them, and normalisation can silently convert them to plain letters.
That last case is the dangerous one: the password you set is not the password that was stored, and you find out at the next login.
02
Mistake two: pasting invisible characters
Copying a password from a document can carry a zero-width space or a non-breaking space. Neither is visible and both are part of the string.
You cannot retype what you cannot see, so the password becomes unusable the moment you are on a device without the clipboard.
✦Type passwords rather than pasting them from documents. An invisible character is a lockout waiting for the wrong moment.
03
Mistake three: predictable substitutions
Replacing a with @ and s with $ adds almost nothing. Password cracking tools have applied these substitutions for decades.
The strength comes from length and unpredictability, not from which characters are used. A long passphrase beats a short one with symbols.
04
Mistake four: ignoring the keyboard you will use
A character easy to type on your laptop may need three taps on a phone or be in a different place on another country's layout.
A password you cannot type on the device you have with you is a password you do not have.
05
Mistake five: assuming every system accepts them
Plenty of systems still restrict passwords to ASCII, and some silently truncate at a length or strip characters they do not like.
Silent handling is the risk. A visible rejection is fine; a silent change means the stored password differs from the one you chose.
- Some systems truncate at a fixed length
- Some strip characters without telling you
- Some normalise Unicode before hashing
06
What to do instead
Use a password manager and let it generate long random strings from ordinary ASCII characters. Length does the work that symbols are imagined to do.
Where you must remember a password, use a long passphrase of unrelated words. It is both stronger and easier to type on any keyboard.
What actually strengthens a password
Comparing approaches by the practical measures that matter: resistance to cracking and whether you can reliably type it.
| Approach | Strength | Reliably typeable |
|---|---|---|
| 8 characters with symbols | Weak | Usually |
| Letter-to-symbol substitutions | Weak | Yes |
| Four unrelated words | Strong | Yes |
| 20 random ASCII characters | Very strong | With a manager |
| Styled or invisible characters | No gain | No |
The bottom row adds no strength and introduces a real lockout risk, which makes it the only genuinely bad option in the table.
07
Copy the symbols discussed in this guide
This guide explains the topic. Use invisible characters for the main copyable set, compare it with punctuation symbols, and browse technical symbols for another useful angle. Each collection includes searchable Unicode names and HTML references.
08
Questions, answered
Can I use emoji in a password?+
Some systems accept them, but support is inconsistent and normalisation can alter the stored value. The risk of lockout outweighs any strength gained.
Do symbols make a password stronger?+
Less than length does. Adding characters to a short password helps far less than making it longer with ordinary characters.
Why did my password stop working?+
A likely cause is an invisible character carried in by a paste, or a system that silently normalised or truncated what you set.
09
Sources and review notes
Recommendations were checked against published platform, Unicode and accessibility guidance. This is a standards review, not a claim of testing every screen reader, app or device combination.
