Why does ChatGPT reply with # and **?

In the previous article, I introduced markdown as the universal language of AI. Now it is time to open a ChatGPT answer and decode every symbol that appears there.

The good news: you do not need to learn markdown deeply. You only need to recognize the symbols to read the answer clearly.

The map of the most common symbols

Think of this as a small dictionary. Read, recognize, move on.

Headings (from chapter to sub-headline)

• # Heading is the largest — equals the page title
• ## Subheading is mid-size
• ### Smaller subheading is the most discreet

The more #, the smaller the heading. Same concept as Word, without clicking any tool.

Emphasis (bold and italic)

• **text** → becomes text in bold
• *text* → becomes text in italic

When ChatGPT wants to flag something important, it uses **bold**. That is how it nudges you: “look at this”.

Lists

• - item → unordered list (bullets)
• 1. item → ordered list (step-by-step)

Numbered lists are especially useful for instructions. ChatGPT loves delivering steps this way: “1. do this, 2. do that”.

Quotes and callouts

• > text → becomes a styled blockquote (the visible “note” block)

Links

• [visible text](https://url) → becomes a clickable link with the visible text

Code (technical block)

• `word` → rendered inline in a gray box
• Three backticks ``` on their own line → opens a code block

The same text, with and without rendering

Imagine ChatGPT gives you a chocolate cake recipe. In raw markdown, what lands on your screen is:

# Chocolate cake
**Ingredients**
- 3 eggs
- 1 cup of chocolate

**Steps**
1. Beat the eggs
2. Add the chocolate

Rendered in any viewer that understands markdown (Word, Docs, Notion, our blog), it becomes:

Same information, completely different reading experience. Markdown is the “barcode” of formatting.

Why this is good for you

• You copy and paste without losing the structure.
• You read anywhere — even in apps that don’t render markdown.
• You edit effortlessly — just adjust the symbols.

And formatting does not solve everything

Worth remembering: AI can deliver a visually perfect answer and still get the content wrong. Formatting well is only half of the work — reviewing, checking numbers, and validating sources is still essential.

For fast lookup at any moment, bookmark the complete Markdown symbol cheat sheet.