How to Fix Weird ChatGPT Spacing in Docs, Word, and a CMS
Why pasted spacing behaves differently
Browse all cleaning guides, or continue here for spacing diagnosis.
Two passages can look alike while containing different source characters. A normal space, non-breaking space, tab, trailing space, and blank line have different layout and validation behavior. The AI spacing fixer exposes those rules separately instead of applying a broad rewrite.
Reproducible destination experiments
Use a new document or draft, paste each source literally or insert the named character, and compare source view or cursor movement before and after cleanup.
| Destination scenario | Source fixture | Expected cleaned text | Rule under test |
|---|---|---|---|
| Google Docs-style no-break copy | Project[U+00A0]status | Project status | normalize-nbsp |
| Word-style tab separation | Name[TAB]Value | Name Value | normalize-tabs |
| CMS repeated spacing | Publish now | Publish now | collapse-spaces |
| Line-ending whitespace | First line[SPACE][SPACE] then Second | No spaces before the line break | trim-trailing-whitespace |
| Excess empty line | One then three line breaks then Two | One blank line between paragraphs | collapse-blank-lines |
The source-to-output pairs are automated engine fixtures. “Google Docs-style,” “Word-style,” and “CMS” describe common paste shapes, not a claim that every version of those products always emits that character.
Preserve layout that carries meaning
Do not normalize tabs in code or tab-separated data without review. Repeated spaces can align plain-text tables. Non-breaking spaces can intentionally keep a number and unit together. Blank lines can be part of poetry or source formatting. Run one rule at a time when the destination is sensitive.
A practical order
Reveal hidden characters first, normalize U+00A0, decide whether tabs are structural, then address repeated and trailing spaces. Finish with blank-line cleanup. See the invisible-character guide for code-point risks, the Markdown removal matrix when markers are mixed with spacing, and the methodology for overlap priority.