TL;DR
ChatGPT outputs Markdown because its interface renders it. The **bold**, *italic*, and ## headings look correct inside ChatGPT but appear as raw symbols everywhere else. Paste into TextPurify and click Clean to strip them all.
You asked ChatGPT to draft a professional email. It came back with solid writing — good structure, right tone, everything you needed. Then you pasted it into Gmail and saw **Strategic Partnership** sitting there in plain text, asterisks completely intact. The same with *italicized phrases* and ## Subject Line headings that appeared as literal hash marks. You are not doing anything wrong, and ChatGPT is not malfunctioning. This is how the tool is designed to work — and understanding why makes it easy to fix.
What Those Asterisks Actually Mean
ChatGPT outputs text in Markdown format. Markdown is a lightweight markup language created in 2004 for writing documents that get converted to formatted HTML. The syntax is simple: **word** renders as bold, *word* renders as italic, and ## Line renders as a second-level heading.
Inside the ChatGPT interface, all of this rendering happens automatically. You see bold text, not asterisks. You see a heading, not a hash symbol. The problem is that every other application in the world — Gmail, Microsoft Word, WordPress, HubSpot, Notion, Mailchimp, Google Docs — does not render Markdown by default. When you copy the text out of ChatGPT, you bring the raw Markdown syntax with it.
Why ChatGPT Is Built This Way
OpenAI trained ChatGPT to output Markdown because the vast majority of its users interact through chat.openai.com, the mobile app, or API clients that render Markdown. Formatting genuinely makes responses easier to read — bold text highlights key points, headings organize long answers, bullet lists make comparisons scannable.
This was a reasonable design decision for a chat interface. The problem is that people use ChatGPT for far more than chatting. They use it to draft emails, write blog posts, create product descriptions, summarize documents, and generate content that ends up in a dozen different destinations — none of which agreed to receive Markdown.
The Full List of Formatting ChatGPT Adds
Asterisks are only the beginning. When you copy a typical ChatGPT response, you might be carrying all of this invisible baggage:
- Double asterisks
**text**— bold markers that appear as literal asterisks in plain-text contexts - Single asterisks
*text*— italic markers - Hash symbols
## Heading— section headings at levels 1–6 - Em-dashes (—) — used as structural separators throughout responses
- Curly / smart quotes ("text") — typographic quotes instead of straight ASCII quotes
- Unicode ellipsis (…) — a single code point that replaces three dots
- Zero-width spaces — invisible characters that cause word-count discrepancies and break find-and-replace
- Backticks
`code`— inline code markers - Hyphen bullet lists
- item— unordered lists
Each of these causes a different kind of downstream problem, depending on where you paste.
Where These Problems Show Up
Gmail and email clients: Asterisks and hash symbols appear as raw characters. Your formatted email looks like markup noise. Curly quotes can encode incorrectly in older email clients, appearing as question marks or boxes.
Microsoft Word: The paste dialog can handle some formatting but not reliably. Bold markers often stay as asterisks. The Unicode ellipsis (…) gets counted differently by Word's word-count tool than three dots would.
WordPress and most CMSs: Markdown formatting is usually stripped or passed through as literal characters. Invisible zero-width spaces corrupt text imports and cause character-count validation to fail in systems with strict field limits.
HubSpot and Mailchimp: These have their own rich-text editors that do not interpret Markdown. Bold **text** stays as asterisks. Em-dashes can cause encoding issues in certain HTML email renderers.
Spreadsheets: Smart quotes break formula strings. Em-dashes misalign text comparisons. Zero-width characters inflate character counts in cells and make VLOOKUP and MATCH functions return false negatives.
How to Remove Asterisks from ChatGPT Text
Manual method: You can run a series of find-and-replace operations — replace ** with nothing, then * with nothing, then ## and ### with nothing, then find each em-dash and each curly quote variant. That is six to ten separate replacements, every single time, for text that may or may not have all of these symbols.
Using TextPurify: Paste your ChatGPT text into TextPurify's ChatGPT cleaner and click Clean. The ChatGPT preset strips asterisks, removes Markdown headings, converts em-dashes to standard hyphens, replaces curly quotes with straight quotes, and removes all invisible Unicode in one pass. Everything runs in your browser — your text never leaves your device.
You can also toggle individual options on or off if you want to keep some formatting. For example, turning off Strip Markdown but keeping Remove Hidden Characters will strip the invisible Unicode without touching the asterisks — useful if you're pasting into a tool that renders Markdown.
Other AI Tools Have the Same Problem
ChatGPT is not alone here. Claude, Grok, DeepSeek, and Jasper all output Markdown by default. Claude adds its own variant with heavy use of bolding and XML artifact tags. DeepSeek adds chain-of-thought reasoning blocks. Grok produces numbered lists with extensive em-dash usage. The cleaning rules are similar across all of them — but TextPurify has dedicated presets for each so you don't have to configure anything manually.
If you publish AI content regularly, building a clean-before-paste habit is the most effective change you can make to your workflow. Five seconds of cleaning prevents formatting bugs that are embarrassing and annoying to fix after the fact.