The job
What the Perplexity zero width space remover is for
A Perplexity answer is almost never the finished artefact. It becomes a paragraph in a brief, a row in a sheet, a block in a CMS or a comment in a file, and the invisible characters it carries only cause damage after that second paste. This tool goes in between: clean the answer once, then move it on.
Fourteen invisible Unicode characters are handled. Eleven are removed outright and three space look-alikes are converted to ordinary spaces so nothing runs together. Nothing else is altered, and the clean runs in your browser, so a confidential research answer is safe to paste.
Before you start
Two things in Perplexity are called a space
Perplexity has a feature called Spaces: named workspaces holding threads, files and instructions. Deleting one happens inside Perplexity, and a web tool cannot reach it.
This page handles the other kind, the character. Zero-width spaces, hair spaces and no-break spaces that come along in the clipboard when you copy an answer out. If the workspace is what you wanted, nothing here will help.
The task
How to remove zero width spaces from Perplexity text in three steps
The right moment is immediately before the paste into your destination, not after that destination has already reformatted the text.
- Step 1. Copy the answer whole
- Citation markers included. The spacing around numbered references is a common place for invisible characters to sit.
- Step 2. Paste and run
- The tool reports how many characters were removed or changed, which tells you whether the answer was carrying anything at all.
- Step 3. Copy the result onward
- Paste the cleaned version into the CMS, the sheet or the repository. The visible wording is unchanged.
The list
Which invisible characters are removed, and which are replaced
Two behaviours, not one, and knowing which is which explains why a cleaned answer is not simply shorter.
- Removed outright
- U+200B zero-width space, U+200C non-joiner, U+200D joiner, U+2060 word joiner, U+FEFF byte-order mark, U+00AD soft hyphen.
- Removed outright, directional set
- U+202A, U+202B, U+202C, U+202D and U+202E. Embeddings, the pop and the two overrides, which reorder text without showing anything.
- Replaced with an ordinary space
- U+00A0 no-break space, U+202F narrow no-break space, U+200A hair space. Deleting these would fuse the words on either side.
- Left alone deliberately
- Joiners inside emoji sequences, because removing those would break the glyph into its parts.
What stays
What this cleaner will not change in your Perplexity answer
Worth reading before you run it, because several tools in this category do far more than their name suggests and hand back text you then have to repair.
Markdown is untouched: headings, bold, lists, links and code fences survive. Punctuation is untouched: curly quotes stay curly, dashes stay as typed, an ellipsis stays a single character. Run spacing and blank lines are untouched, so the shape of the answer on the page is the shape you get back. If you want any of those changed, that is a different tool and this one will not do it behind your back.
The cause
Why Perplexity output carries more of this than a plain chat reply
Perplexity answers by fetching pages, quoting them and summarising them. Published pages are full of deliberate invisibles: a narrow no-break space before a unit so it cannot wrap, a soft hyphen at a break candidate inside a long term, a byte-order mark written by an exporter.
That material gets carried into an answer, rendered as styled text, and copied into your clipboard with the invisibles intact. A model answering without a retrieval step has one source of residue. A retrieval engine has two, and the second belongs to somebody else's typesetting.
The model picker adds a third variable. Perplexity lets you choose which model answers, so two sessions on the same account are not guaranteed to produce the same residue.
The cost
What a zero width space breaks once the text is pasted
Everything here shows up one step after the paste, which is exactly why it takes so long to find. A zero-width space in a heading splits the keyword, and the slug your CMS generates stops matching the visible title. A no-break space in a formula turns a number into a string.
Further downstream, find-and-replace misses obvious matches, a CSV import rejects a column that looks perfectly normal, a screen reader pauses in the middle of a word, and a diff reports a changed line where nothing visible changed.
Straight answer
Does cleaning zero width spaces hide that you used AI?
No. Detectors read sentence structure, word choice and rhythm, and none of those move when an invisible character is deleted. Text that was flagged before the clean reads the same afterwards.
The reason to do it is mechanical, and it is a good enough reason on its own. Invisible characters break the systems your text is about to enter. Removing them fixes that. It does not change what the writing sounds like, and nothing on this page will claim otherwise.