What is different about Perplexity output?
Perplexity is a search-based assistant, so its answers come with numbered source references. Those references appear inline as [1], [2] and so on. When you copy the answer, they can come along as plain text next to the words they support.
| Artifact | Removed by the cleaner? | How to handle it |
|---|---|---|
| [1] [2] citation markers | No | Regex find and replace |
| Markdown symbols | Yes, with Strip Markdown on | Markdown to Text preset |
| Curly quotes, em dashes | Yes, in AI Clean | Typography rule |
| Non-breaking or narrow spaces | Yes | Exotic spaces rule |
Step by step
- Copy the answer and paste it into the cleaner above.
- Keep the AI Clean preset and switch on Strip Markdown if the destination is plain text (see Markdown to plain text).
- Copy the result and, if you want the citation numbers gone, replace \[\d+\] with nothing in any editor that supports regular expressions, then collapse any double spaces.
The em dashes and curly quotes that assistants favour are explained in normalizing quotes and dashes and removing em dashes from AI text.
Think before deleting the citations
The numbers point to the pages the answer drew on. If you delete them, keep the URLs and verify the claims you are relying on. An unsupported statement is easier to spot when its source is attached, and easier to publish by mistake when it is not.
What cleaning does not do
It does not remove statistical watermarks, which change word choice and add no characters, and it does not change how AI detectors score text. Detectors are explained in how AI detectors work, and the general cleaning workflow is in clean AI text.