A round trip is not an identity operation

Convert Markdown to HTML, convert it back, and you do not get your file. You get a document that renders identically and reads identically — and whose text is different. The differences are systematic rather than random, which means you can know in advance exactly what will change and decide whether that is acceptable.

The pipeline is marked going out and turndown coming back. At each stage information is either recorded or discarded, and the discarded parts do not return.

Syntax choices collapse to one style

Markdown allows several ways to write the same thing. Headings by hash prefix or by underline. Emphasis with asterisks or underscores. Bullets with hyphens, asterisks or plus signs. Soft and hard line breaks.

HTML records only that something is a heading, or emphasis, or a list item — not which syntax produced it. So the return trip writes the converter's single preferred form throughout: hash headings, asterisk emphasis, hyphen bullets, fenced code.

The consequence is the practical one: if the original mixed styles, which is normal in a file several people have edited, every affected line comes back rewritten. In version control that reads as a mass rewrite with no change in meaning, and it buries whatever you actually edited.

Comments do not come back

An HTML comment in the original passes through to the HTML stage and is not reproduced on the way back. A <!-- TODO --> note you left for yourself is simply gone from the result.

Reference links become inline

Markdown has two link syntaxes: inline, and reference-style with the definitions collected at the bottom. HTML has one — an href. So every reference link returns as an inline link and the definition block disappears.

For a document written deliberately in reference style, so that the prose stays readable while editing, this undoes a formatting decision that was made on purpose.

Whitespace normalises

Hard breaks written as two trailing spaces become explicit break elements in HTML and come back as plain newlines. Multiple blank lines collapse. Indentation inside list items is regularised. The rendered result is identical; the raw file is tidier and different.

Raw HTML survives, sometimes reshaped

HTML embedded in your Markdown passes through, and on the return journey is converted back to Markdown where an equivalent exists or left as HTML where none does — occasionally with different escaping than you wrote.

The rule: meaning yes, text no

The round trip preserves meaning reliably and preserves text not at all. That single sentence tells you when to use it.

It is unsafe in three situations. Under version control, where formatting-only diffs hide real edits. Where a house style exists, since the converter's defaults replace it wholesale. And where other people will edit the file afterwards, because they inherit the reformatting and the churn repeats on every pass.

Where it is exactly the right tool

Two cases, and in both the normalisation is the goal rather than a side-effect.

Standardising a directory of Markdown written by many hands: run every file through, commit the result as a single deliberate style pass, and everyone works from one format afterwards. The mass diff is the point, and doing it once beats arguing about it forever.

Recovering a document that has accumulated stray HTML over years of edits: converting out and back leaves clean, consistent Markdown.

Before you overwrite anything

Convert a copy and compare it against the original with a diff. If the only changes are the style normalisations described above, proceed. If content is missing or a link looks malformed, stop — that is a genuine failure rather than a formatting one, and it is much easier to spot now than after the original is gone.

Everything runs in your browser and nothing is uploaded. For the single direction without the return journey, see the HTML to Markdown converter.