This conversion is lossy on purpose

HTML is a rich document format — structure, attributes, classes, inline styles, arbitrary nesting. Markdown has a small vocabulary: headings, emphasis, lists, links, images, code, blockquotes, and in extended flavours, tables.

Converting down means everything outside that vocabulary is dropped or approximated. That sounds like a criticism and it is not: the loss is usually why you are doing it.

What survives

The conversion uses turndown, and the mapping is what you would expect: heading elements become hash-prefixed headings, paragraphs become paragraphs, strong and bold become double asterisks, em and italic become single asterisks, anchors become bracketed links with the URL in parentheses, images become the same with a leading exclamation mark, list items become hyphen or numbered lists, and blockquotes become lines prefixed with a chevron.

Code is worth separating: an inline code element becomes backticks around the text, while a preformatted block becomes a fenced block. Those are different constructs and conflating them is a common source of confusion when the output looks wrong.

What is lost, and why that is the point

• Classes, IDs, inline styles and data attributes vanish. Markdown has nowhere to store them.
• Layout containers disappear. Nested divs and spans carrying no meaning are precisely the thing you are trying to shed.
• Anything with no Markdown equivalent — form controls, iframes, custom elements — either drops out or comes through as raw HTML.

Here is the standard to judge it by. As a faithful copy of the HTML, the conversion fails and always will. As a way to recover the content from the formatting, it succeeds, and that is what people actually want when they paste a published page into a documentation repository. Measuring it against the wrong standard is why people conclude the tool is broken when it did exactly the right thing.

Where to check the output

• Tables. Simple ones convert to Markdown table syntax. Merged cells, nested elements or anything irregular will need a hand.
• Nested lists. They survive, but unusual nesting can come out indented oddly.
• Formatting that crosses element boundaries. Emphasis split across tags in awkward ways gets simplified.

Why people do this

• Moving a published article into a documentation repository or a static site
• Migrating content between systems while keeping the text and its structure
• Cleaning text pasted out of a word processor, which arrives wrapped in generated markup
• Getting content into a form that reviews well in a pull request

That last one is underrated: Markdown diffs readably and HTML does not, so converting is often what makes content reviewable at all.

Everything runs in your browser and nothing is uploaded. Input up to 5 MB is accepted. To go the other way, or to understand what happens when you convert both ways, see the round-trip converter.