Does ChatLab support Markdown in training data?

Yes, you can use Markdown syntax in Text and Q&A training sources to add structure, headings, lists, and links. Website sources are automatically converted to Markdown format during scanning.

How much extra space does Markdown formatting use?

Markdown formatting uses about 30% more training characters than plain text because the syntax characters (such as #, **, [], etc.) are included in the character count.

Will my chatbot format its responses with Markdown?

Yes. When your training data includes Markdown, the chatbot is more likely to structure its responses with headings, lists, bold text, and clickable links in the chat widget.

Which Markdown elements are supported in chat responses?

The chat widget renders headings, bold text, italic text, ordered and unordered lists, links, images, and superscript. Tables in training data are understood by the AI but may be rendered as lists in responses.

Advanced training using Markdown syntax

Markdown is a lightweight formatting syntax that adds structure to plain text. When you provide training data with Markdown, the AI can better understand sections, hierarchies, and references in your content -- and it will use that structure when responding to users.

How Markdown works in ChatLab

Markdown is used in two ways depending on the training source:

Website sources -- Markdown is applied automatically. When ChatLab scans a website, it converts the HTML into Markdown format, preserving headings, lists, links, and page structure.
Text and Q&A sources -- You write Markdown manually. The text fields accept raw Markdown syntax, which is stored as-is and used during training.

In both cases, the structured data helps the AI recognize sections, distinguish key information, and include relevant links in its answers.

Markdown in website sources

When training from websites (Training > Links), ChatLab automatically converts the page content to a text format that preserves links. External links are stored using Markdown link syntax, so the AI can reference them in responses.

To view how a scanned page is stored, click the View button next to any trained link. The preview shows the stored content, including any Markdown-formatted links.

In the example above, the email address is stored as [info@yoursite.com](mailto:info@yoursite.com) -- a Markdown link that the AI can include as a clickable link in its responses.

Markdown in Text sources

Navigate to Training > Text and click Add New Training: Text. In the Text field, you can type Markdown syntax directly.

The example above shows a return policy written with Markdown headings (#, ##), bold text (**30 days**), bullet lists (-), and links ([Returns Portal](https://example.com/returns)).

Markdown in Q&A sources

Navigate to Training > Questions & Answers and click Add New Training: QA. The Answer field supports Markdown syntax, allowing you to structure responses with lists, bold text, and links.

Supported Markdown syntax

Here is a reference of the Markdown elements you can use in Text and Q&A training sources:

Headings -- Use # for a main heading, ## for a subheading, down to ###### for the smallest heading level.

# Main heading
## Subheading
### Smaller heading

Bold and italic -- Wrap text in **double asterisks** for bold or *single asterisks* for italic.

**bold text**
*italic text*

Unordered lists -- Start each line with - or * followed by a space.

- First item
- Second item
- Third item

Ordered lists -- Start each line with a number followed by a period and a space.

1. First step
2. Second step
3. Third step

Links -- Use the format [link text](URL) to create clickable references.

Visit our [Help Center](https://example.com/help)
Email us at [support@example.com](mailto:support@example.com)

Images -- Use ![alt text](image URL) to reference images. The AI may describe or link to these in responses.

![Product photo](https://example.com/images/product.png)

How Markdown appears in chat responses

When your training data is structured with Markdown, the chatbot renders formatted responses in the chat widget -- including bullet lists, bold text, numbered steps, and clickable links.

The chat widget supports rendering of headings, bold text, italic text, lists (ordered and unordered), links, images, and superscript.

Character usage

Markdown syntax characters (#, **, -, [](), etc.) count toward your training data character limit. Markdown-formatted content typically uses about 30% more characters than the equivalent plain text.

For example, the link [Help Center](https://example.com/help) uses more characters than just writing "Help Center", but gives the AI a URL it can include in responses.

If you are close to your character limit, you can reduce usage by simplifying the Markdown -- for instance, using fewer headings or shortening link labels.

Tips for effective Markdown training

Use headings to organize topics -- This helps the AI identify which section of your content is relevant to a user's question
Use lists for options, steps, or features -- The AI will often mirror this structure in its responses
Include links to important pages -- The AI can provide these links when answering related questions
Keep it simple -- Basic Markdown (headings, lists, bold, links) is most effective. Complex formatting like nested tables may not improve responses

Adding plain text -- Add text content without Markdown
Adding questions and answers -- Define Q&A pairs for precise responses
Training your chatbot: basics and limits -- Understanding training data limits
Understanding training data limits -- How character counts work