HTML to Markdown Converter

Convert HTML to GFM Markdown and see what was left out

This converter turns the headings, paragraphs, lists, tables, code and links of an HTML document into GitHub Flavored Markdown (GFM). <h2> becomes ##, <strong> becomes **, a link becomes [text](url), and a table becomes a Markdown table with its columns split by |. Unlike a tag stripper, it keeps the structure of the document, so the result can go straight into a GitHub README or a notes app that reads Markdown.

You can paste HTML code, or paste text you copied from a web page or Google Docs. Copied text arrives with the HTML formatting the browser puts on the clipboard, and that is what gets converted. Below the result, a structure map shows the order of blocks in the document, and a conversion log lists what Markdown could not carry as it was, such as merged cells, relative URLs and embedded videos.

When you need to turn HTML into Markdown

• Moving a blog or CMS post elsewhere: CMS posts, WordPress included, are stored as HTML. Converting them is faster than retyping headings, lists and links when you move to a static site generator or a Markdown notes app.
• Turning a web page into a README or developer doc: paste a page with install steps and the code block languages (language-bash) and task-list checkboxes come through as GFM.
• Moving Google Docs or meeting notes into Markdown notes: text copied from a document editor often writes bold and italic as style values instead of tags. The converter reads those values and writes **bold** and *italic*.
• Feeding a web page into an AI tool or a note in less space: without class and style attributes and scripts, the same content is much shorter. In the examples above, "Copied from a document editor" shrinks from 1,057 bytes to 217, and "Newsletter email" from 997 to 362.
• Archiving a newsletter or email: email HTML is full of layout tables and 1×1 images that report when it was opened. Layout tables are unwrapped into text and tracking images are dropped, which leaves a readable document.

How to use it

1. Add your HTML: paste code into the HTML box, or select the part of a web page you want, copy it (Ctrl+C) and paste. Open an HTML or HTM file with "Open file" or drop it onto the box. While an example is loaded, pasting replaces it.
2. Check how it was pasted: text copied from a page goes in as HTML formatting, while HTML code copied from a code editor goes in as plain characters. If it picked the wrong one, click "Paste as plain text" or "Paste with formatting" in the status line.
3. Choose a Markdown style: heading marker (# or underline), list bullet (- * +), emphasis marker, code fence, link style and table format. The result is rewritten as soon as you change a setting.
4. Set a base URL: if the HTML has relative URLs such as /images/a.jpg, enter the address of the original page in the base URL box to make them full URLs. A base, canonical or og:url address in the HTML is filled in automatically.
5. Review and take it: check the preview, click a band of the structure map to jump to that part of the result, and read the conversion log. Use "Copy Markdown" or "Save .md". Ctrl+Enter (⌘Enter) in the result box also copies.

How each tag is converted

These are the results the converter actually produces with its default settings (headings #, bullets -, emphasis *, fence ```, inline links, Markdown tables).

Where Markdown breaks in everyday English text

Markdown symbols work or fail depending on the characters around them, so plain text that merely looks harmless can change meaning. The "Written with plain symbols" column shows what happens when you only swap the tags for symbols; each one was rendered with markdown-it (CommonMark 0.31.2) and marked (GFM) to check. The converter adds a backslash only where one is needed. Languages that attach particles to words, such as Korean and Japanese, hit a further problem with ** next to punctuation, which the converter handles by writing <strong>.

The last row is deliberate: a backslash that is not needed only makes the Markdown harder to read, so the converter leaves it out. The same goes for a # that is not at the start of a line, as in C#. If you pick _ as the emphasis marker, only the spots where it would not work inside a word, such as un__break__able, are written with ** instead.

Pasting from a web page or Google Docs

When you copy text in a browser, the clipboard holds both the visible characters (text/plain) and the formatted version (text/html). The converter reads both at the moment of pasting. If the plain text visibly contains tags, it treats the paste as HTML code; if not, it treats it as copied page content. So dragging across a web page brings in that part's HTML, while copying HTML code from a code editor brings in the code characters rather than the editor's syntax colors.

Google Docs wraps the whole copied range in <b style="font-weight:normal" id="docs-internal-guid-…"> and writes real bold and italic as font-weight:700 and font-style:italic on spans. Several editors have been reported to read that wrapper b tag as bold, so everything you paste ends up wrapped in **…**. This converter does not treat a b tag with font-weight:normal as bold, and carries bold, italic, strikethrough and superscript over from the style values of the spans inside. The example "Copied from a document editor" above is exactly this kind of HTML.

Things to know when converting tables

A GFM table needs a header row first, then a delimiter row of ---. The colons in the delimiter row set alignment: :--- is left, ---: is right and :---: is centered. The converter takes alignment from the align attribute or style text-align, and pads the cells with spaces so the columns also line up in the raw Markdown (a column wider than 40 characters is not padded).

• A | inside a cell: written as \| so it is not mistaken for a cell border. Per the GFM spec, a | inside code or a link must be written the same way.
• Several lines in one cell: GFM table cells cannot hold blocks such as paragraphs or lists, so the lines are joined with <br>.
• Merged cells: a Markdown table cannot merge cells. By default the merged spots are spread out as empty cells; if you switch the table setting to HTML, the table stays as an HTML table with its rowspan and colspan, minus class and style.
• Tables without a header: since a header row is required, the first row is used as the header. The conversion log reports how many.
• Layout tables: a table with role="presentation", a table inside a table, or a table with a single column is treated as page layout rather than data, and its cells are unwrapped into text one after another. This is common in email HTML.

Choosing a Markdown style

How much smaller does it get?

Removing styling attributes, scripts and layout tables cuts the size a lot, even though the content is the same. These are the values the size bar shows when you open each example (bytes in UTF-8). Document-editor HTML, which is full of style values, and email HTML shrink the most.

Common mistakes

• Pasting the whole page source and getting the menus too: view-source code includes the top navigation and footer links. If you only need the article, select just that part on the page, copy it and paste.
• Publishing relative URLs as they are: /uploads/a.jpg only opens on the original site. Enter a base URL to turn it into a full URL.
• Thinking the images are missing from the preview: the preview does not load images, for privacy and speed, and shows an "Image" marker with the alt text instead. A ![description](url) in the result is correct.
• Reading an empty cell in a Markdown table as a missing value: it may be a spot where a merged cell was spread out. If the log mentions merged cells, compare with the original table or keep the table as HTML.
• Deleting the two spaces at the end of a line: a line break (br) is written as two trailing spaces. An editor that trims trailing whitespace on save will remove the line breaks.
• Pasting into a place that does not accept HTML tags: superscript, keys, collapsible sections, and HTML tables are written as tags. In an editor that strips HTML, the text stays but the formatting is lost.

Limits

• Markdown describes document structure, so CSS color, fonts, spacing and layout are not carried over.
• It does not fetch a page from a web address (URL). Content that only appears after logging in, or that a script fills in later, has to be copied from the screen and pasted.
• Image and video files do not go into the .md; only their URLs remain. Images embedded in the page as data: URLs are dropped because they make the result far too long.
• iframe, video and audio remain only as links to their URLs, and forms (input fields, buttons) and SVG graphics are not carried over.
• HTML from a word processor that writes lists as paragraphs starting with a bullet character may come out as paragraphs, not a list.
• Input is read up to 2 MB. How the result looks on GitHub follows the GFM spec; editors differ in how they handle HTML tags and line breaks.

Sources

Basis: Markdown syntax comes from the GFM 0.29-gfm and CommonMark 0.31.2 specifications, and browser behavior from MDN. All were checked on 22 October 2026. The renders in the table above were produced with markdown-it (CommonMark) and marked (GFM), and the Markdown in the tables is this tool's own output.

• 01GitHub Flavored Markdown Spec 0.29-gfm4.10 Tables (delimiter row, alignment, \| inside a cell even in code, example 200, no blocks), 5.3 Task list items, 6.5 Strikethrough (one or two tildes, example 491), 6.11 Disallowed raw HTML
• 02CommonMark Spec 0.31.2Left- and right-flanking rules for emphasis, backslash escapes, ordered list numbers (up to 9 digits), hard line breaks
• 03MDN — DOMParser.parseFromString()Scripts in a document parsed as text/html are not executed
• 04MDN — Element: paste eventReading pasted content with clipboardData.getData() and changing the default behavior
• 05RFC 7763 — The text/markdown Media TypeThe media type name for .md files, text/markdown (March 2016)
• 06MarkText #4688 — Pasting from Google Docs makes the whole paragraph boldThe Google Docs docs-internal-guid b wrapper and span font-weight values

Frequently asked questions

Tools to use with this one