The asterisks, pound signs and dashes in AI answers are Markdown, a simple formatting code that chatbots write in. Inside the chat app, **word** shows up as word and ## Heading shows up as a heading. When you copy the raw text into something that doesn’t understand Markdown, like a text message, a web form or some email apps, you see the symbols instead of the formatting. The quickest fix is to ask for plain text, and there are a few other tricks for when you’ve already got the answer.
What do the symbols mean? #
| You see | Markdown meaning | Shows up in the chat app as |
|---|---|---|
**text** | Bold | text |
*text* or _text_ | Italic | text |
#, ##, ### at the start of a line | Headings, from largest to smallest | A heading |
- or * at the start of a line | A bullet point | • item |
1. at the start of a line | A numbered list | 1. item |
`text` | Inline code | A code-styled word |
| Three backticks on their own line | A code block | A shaded box, often with a copy button |
| Lines with ` | pipes and—` dashes | A table |
[words](link) | A link | Clickable words |
Why do AI chatbots write in Markdown? #
Language models learn from huge amounts of text that includes documentation, forums and code repositories, where Markdown is everywhere. It’s also a practical choice: Markdown is plain text that any app can store and send, and chat apps can turn it into clean formatting on screen. Headings, bullets and bold make a long answer easier to scan, so chatbots are tuned to use them.
The problem is only what happens after you copy.
Where the asterisks show up #
- Text messages and most messaging apps. SMS and iMessage don’t render Markdown, so you get the raw symbols.
- WhatsApp. It has its own formatting where single asterisks mean bold, so an AI’s
**bold**comes out with stray extra asterisks. - Web forms, job applications and support tickets. These usually take plain text only.
- Some email apps, depending on how you paste.
- Word processors, unless they have a Markdown option switched on.
Places that do understand Markdown include notes apps built on it, such as Obsidian, and sites like GitHub, Reddit and Discord. Pasting there keeps the formatting.
6 ways to get clean text from an AI #
- Ask for plain text. Add “Reply in plain text. No Markdown, no asterisks, no pound signs, no bullet points.” This works in every chatbot.
- Make it the default for a chat. If you use one chat for texts and emails, put the plain-text instruction in that chat’s system prompt so you don’t repeat it. Our system prompt examples include one for writing messages.
- Ask for the format of the destination. “Write this as a text message,” “as an email body,” or “as a WhatsApp message using WhatsApp’s formatting” gets you something ready to paste.
- Swap tables for lists. Tables turn into rows of pipes when pasted. Ask for “a simple list instead of a table.”
- Copy from the rendered text. Some apps copy the formatted version when you select text on screen rather than using a “copy raw” option. Try selecting just the sentences you need.
- Find and replace. In a word processor, replace
**with nothing, then##and#with nothing. It takes seconds for a long document.
When you want to keep the formatting #
Markdown is useful when the destination understands it. Code is the clearest case: keep code blocks intact and use the block’s copy button so indentation survives. Notes in a Markdown app, posts on forums that support it, and documentation all benefit from keeping the symbols.
What good formatting looks like in a chat app #
A chat app should render Markdown properly and make copying easy, since most AI answers end up somewhere else. Personal LLM renders code blocks with a copy button, tables and lists, and lets you copy a whole reply or just the sentence you need in one tap. You can give each chat its own system prompt, so a “texting” chat always answers in plain text while a “study notes” chat keeps headings and bullets. It runs open models such as Qwen 3.5 and Gemma 4 on the phone itself, so drafts of personal messages aren’t uploaded anywhere, and it works offline once a model is downloaded.
If you often use AI to polish what you send, see our guides to checking a message’s tone and writing a cover letter with AI, both of which end in text you’ll paste somewhere without Markdown.
Frequently asked questions #
Why does ChatGPT put ** around words? #
Double asterisks are Markdown for bold. The ChatGPT app displays them as bold text, but if you copy the raw text into an app that doesn’t support Markdown, the asterisks appear instead of the formatting.
How do I copy AI text without the asterisks? #
Ask the chatbot to reply in plain text with no Markdown before you copy, or select and copy the displayed text rather than the raw version. If you’ve already pasted it, use find and replace to remove the ** and # symbols.
What is Markdown? #
Markdown is a lightweight way to format plain text using symbols: asterisks for bold and italics, pound signs for headings, dashes for bullets and pipes for tables. Apps that support it turn the symbols into formatting, and apps that don’t just show the symbols.
Why do AI answers have ### in them? #
Pound signs at the start of a line are Markdown headings. One is the biggest heading, and three is a smaller subheading. They’re meant to be displayed as titles, not shown as symbols.
Can I stop AI from using bullet points and headings? #
Yes. Tell it to write in plain paragraphs with no headings or bullet points. To make it stick for a whole conversation, put that instruction in the system prompt or your custom instructions.