How to use this tool
Use the analyzer to inspect selected unusual characters without changing your input. It is useful when a copied value looks right but fails a comparison, when spacing seems inconsistent, or when you need a code point to include in a bug report. The analyzer is deliberately an inspection tool. Pressing Analyze text produces an unchanged copy, and the report explains supported findings rather than automatically deciding that those characters should be removed.
Scroll the table sideways to see every column.
| Output | Meaning | Does not mean |
|---|---|---|
| Special-character count | Findings in supported categories | Number of errors or AI markers |
| U+ code | Character identifier | Origin or author |
| Position | One-based code-point location | Byte offset or visible glyph count |
| No special characters | No matches in supported categories | Complete Unicode or security validation |
What the analysis covers
The scanner recognizes a selected set of invisible characters, special spaces, em-dash variants, and control or directional marks. Examples include U+200B, U+00A0, U+2014, and U+202E. It does not provide an exhaustive character inventory, a Unicode-property database, or a security certificate for the passage. Its usefulness depends on matching this scope to the problem you are investigating.
Ordinary Latin letters, many accented characters, and most other writing-system characters do not appear as findings. Ordinary spaces, tabs, and standard line breaks are also not special findings. The preview controls can visualize spaces and tabs, which helps examine layout without treating normal text structure as an error. An empty report means no supported special characters were detected; it says nothing about authorship or factual accuracy.
Read a code point and position
A code point is a numeric identifier displayed in the report as U+ followed by hexadecimal digits. The name beside it is a useful explanation, while the code is the more precise detail to record. Some controls receive a general category label instead of a specific formal name. Include both the code and the nearby text when describing an issue to another person.
Locations are counted by Unicode code points and displayed starting at one. They are not byte offsets, and they are not necessarily the same as an editor's column numbers. For example, a joined emoji can contain several code points while appearing as one visible symbol. If a reported location seems surprising, reduce the sample and keep the symbols before the finding in view rather than assuming the counter is measuring screen width.
Try the example and inspect its limits
The example contains café, a tab, and an actual U+200B before the word marks. The accent is ordinary content for this scanner and is not flagged. Show tabs reveals the tab in the preview, while the hidden separator produces a finding. Press Analyze text and compare the result: the original sequence is preserved because this page does not apply removal rules.
This example is useful precisely because not everything unusual to an English keyboard is an issue. If your task is to inspect every character in café or compare precomposed and decomposed accents, you need a full sequence inspector. noclau does not perform general normalization or display every combining mark. Use the report for its supported categories rather than treating it as a complete Unicode debugger.
Investigate an exact-match failure
Keep two copies of a failing value: the expected string and the pasted string. Inspect the pasted one for supported findings. If an invented value AC[U+200B]42 contains a hidden separator, the finding can suggest why it differs from AC42. The bracketed label here represents one real character. Recording the value this way makes the difference visible in a support message.
Next, use the appropriate cleaner on a working copy and rerun the comparison in the original application. Change one category at a time so the result remains interpretable. If cleanup does not resolve the mismatch, inspect ordinary leading spaces, case, line endings, normalization, and visually similar letters. The analyzer cannot infer which differences a receiving application chooses to ignore or reject.
Interpret findings as evidence, not instructions
A U+200D inside an emoji can be expected. A no-break space in a layout may be intentional. A directional mark in multilingual text may need to remain. The report describes the underlying sequence; deciding whether a mark is appropriate requires its context. For text in a language you do not read, obtain a language review before changing joining or direction behavior.
If you move to a remover, start with a limited rule and review the diff. Conservative cleanup preserves several joiners but still removes some potentially intentional characters. Directional and control marks are reported and preserved for manual review. An unchanged result on this analyzer page is always expected. It is not a sign that the detected character escaped a hidden removal stage.
Handle long documents with a smaller sample
The report counter includes all supported findings, but the visible badges show only the first thirty. The text and diff previews also have display limits. For a large document, isolate the paragraph or field that reproduces the issue and inspect it separately. This makes locations easier to follow and avoids drawing conclusions from a preview that omits a later finding.
Keep enough surrounding text to preserve the behavior. A direction issue may disappear if you reduce a mixed-language paragraph to one character. Conversely, a simple identifier mismatch may need only a few letters. An effective minimal sample is small enough to understand and complete enough to reproduce the operation that fails in the destination.
Copy, privacy, and a useful next step
Copy result becomes available after analysis when the unchanged output is nonempty. If clipboard access is blocked, select and copy the result manually as the tool advises. The preview's dots and arrows are not copied. Pasted text is processed in browser memory by the application and is not uploaded for analysis. Website delivery still produces ordinary network requests covered by the privacy policy.
When reporting a problem, provide the code point, displayed position, selected page, browser, and expected outcome. Use an invented sample when possible. The best next step may be a targeted edit, a destination setting, a full Unicode comparison, or no change at all. Inspection is successful when it gives you the information needed for that decision, even if the final text retains every detected character.