Find zero-width, bidirectional and control characters with their line and column, and characters that look like different letters, then clean either one.
Waiting for text
Finds characters you cannot see and characters that look like other characters. Nothing is uploaded.
The findings appear here
Every invisible and control character with its line and column, and every character matching a known lookalike.
Invisible Character Inspector runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
Why does this string not match has two common answers that are invisible on screen: a character you cannot see, or a character that looks like a different one. This finds both from one paste, because working out which of the two you have is the hard part.
Finding one is not evidence of anything, and that is the thing this page is careful about. A zero-width joiner is how a multi-person emoji is assembled. A non-breaking space is in almost every document typed on a Mac. A right-to-left mark is correct and necessary in Arabic and Hebrew text. A variation selector is how an emoji chooses its presentation. The tool reports what is present and where; it does not decide what it means, and a report full of findings is an ordinary result for ordinary international text.
One finding does deserve a closer look: a directional override that is opened and never closed. After it, text renders in an order that does not match the order it is stored in. In prose that is usually a copy-paste that lost its closing mark. In source code it is the mechanism behind a class of attack where the code a reviewer reads is not the code the compiler sees, so the tool flags the imbalance specifically rather than counting the characters and moving on.
The lookalike table is a curated subset and the page says so plainly. Unicode's full confusables data runs to tens of thousands of mappings; this carries the Cyrillic, Greek and fullwidth lookalikes that actually turn up in spoofed domains and pasted identifiers. A clean result means nothing in that subset was found — not that the text contains no lookalikes, which is a claim this tool is not in a position to make.
There are two separate cleaning actions because they fix two different problems and combining them would hand back a string altered in ways nobody asked for. Removing invisibles strips zero-width, bidirectional, tag and soft-hyphen characters and leaves tabs, newlines and non-breaking spaces alone, since those are usually deliberate. Substituting lookalikes rewrites letters, which is only correct if you already know the text was meant to be Latin.
No. A zero-width joiner builds an emoji, a non-breaking space is in most documents written on a Mac, and a right-to-left mark is necessary in Arabic and Hebrew. The tool reports what is there; deciding what it means needs context it does not have.
A character that switches text direction with no matching character to switch it back. The text after it renders in an order different from the order it is stored in. In prose it is usually a lost closing mark; in source code it deserves careful reading.
It is a curated subset covering the Cyrillic, Greek and fullwidth characters that appear in real spoofing. Unicode's full confusables data is far larger, so a clean result here is not a guarantee that the text contains no lookalikes.
Because they do different things. Removing invisibles is safe and reversible in effect. Substituting lookalikes rewrites letters, which is only right if you know the text should be Latin — applying both at once would change more than you asked for.
No. Those are notable characters and almost always deliberate, as is a non-breaking space. The clean action removes zero-width, bidirectional, tag and soft-hyphen characters only.
No. Everything runs in your browser and nothing reaches a server or analytics, which matters because the string being inspected is often an identifier or a credential.