How to Remove ChatGPT Formatting Without Rewriting the Text
Start with the formatting, not the prose
Browse all cleaning guides, or continue here for the formatting workflow.
Formatting cleanup and rewriting are different jobs. The local formatting remover applies selected deterministic rules in your browser. It does not ask a model to improve the wording. Start with safe spacing and hidden-character rules, then enable Markdown removal only when the destination needs plain text.
Six tested paste samples
| Artifact | Before | After | Rule |
|---|---|---|---|
| Heading marker | ## Launch plan | Launch plan | remove-markdown-headings |
| Emphasis marker | This is **important**. | This is important. | remove-markdown-emphasis |
| Code fence | A txt fence around alpha | alpha plus its line break | remove-markdown-code-fences |
| Markdown link | Read [the method](/methodology). | Read the method. | remove-markdown-links |
| Non-breaking space | Price:[U+00A0]$10 | Price: $10 | normalize-nbsp |
| Mixed artifacts | ## Read [this](/privacy) now | Read this now | heading, link, and repeated-space rules |
These fixtures run against the production cleaning engine in CI. The link rule preserves visible anchor text but discards the destination. The fence rule preserves fenced contents but removes the delimiter lines.
A review-safe workflow
Paste the text, inspect detected categories, leave destructive rules off unless needed, run the cleanup, and review the visible change list before copying. Undo restores the original browser-state input. If the wording itself needs revision, use the AI humanizer separately and account for its server and model-processing boundary.
What cleanup cannot preserve automatically
Plain text cannot retain rich-text font sizes, colors, comments, tracked changes, or clipboard-only metadata. Removing Markdown links intentionally loses their URLs. Removing heading markers also loses the encoded heading level. Code, tables, bidirectional text, and language-specific shaping deserve manual review.
Related guides and evidence
Use the Markdown removal matrix for element-level behavior, the invisible-character guide for Unicode evidence, the weird-spacing guide for destination-specific diagnosis, and the cleaner-versus-humanizer comparison when wording is the issue. Full rule definitions and risk labels live in the methodology.