HTML Table to Markdown & CSV

Getting the table out of a page

You rarely have the HTML of a table sitting in a file. It is on a pricing page, in a wiki, in a Confluence export, or in an email. The quickest route is the browser's developer tools: right-click any cell, choose Inspect, walk up the tree to the <table> element, right-click it and choose Copy → Copy outerHTML. Paste that here.

Pasting a whole page works too. Every top-level table in the document is listed, with its caption when it has one, and you pick the one you want. Nested tables (a table inside a cell) are folded into the parent cell as text, because neither Markdown nor CSV can represent them.

<table>
  <tr><th>Plan</th><th>Price</th></tr>
  <tr><td>Free</td><td>$0</td></tr>
  <tr><td>Pro</td><td>$12</td></tr>
</table>

| Plan | Price |
|:-----|:------|
| Free | $0    |
| Pro  | $12   |

How the header row is found

A row is treated as the header when it lives in a <thead>, or when every cell in the first row is a <th>. Many real tables use neither (a first row of plain <td> cells styled bold with CSS), so the detection is only a default. The First row is header box overrides it either way.

GFM tables must have a header row. When you switch it off, the output keeps an empty header line above the alignment row so the table still renders; the first data row is not promoted. CSV has no such rule, and simply emits every row.

What happens to colspan and rowspan

Neither Markdown nor CSV has merged cells, so a spanning cell has to be flattened into a rectangle. Two strategies are offered:

• Repeat value copies the cell's text into every position it covered. Choose this for data you will sort, filter or load into a spreadsheet: each row stays self-describing.
• Leave blank keeps the text once, in the top-left position, and leaves the rest empty. Choose this when the output is for reading and you want it to look like the original.

<tr><td rowspan="2">Team</td><td>Seats</td></tr>
<tr><td>Projects</td></tr>

| Team | Seats    |
| Team | Projects |

A warning under the output tells you when merged cells were expanded, so you know the grid is wider or taller than the visible table was.

Bold, code, links and line breaks

With Keep bold, code & links on, <strong> becomes **bold**, <code> becomes a backtick span, and an absolute link becomes [text](https://…). Relative links keep their text only, because a Markdown file has no base URL to resolve them against. Images are replaced by their alt text. Turn the option off for plain text in every cell.

Line breaks inside a cell (<br>, or several <p> elements) are a problem for GFM, whose rows are single lines. They are written as a literal <br>, which GitHub, GitLab and most renderers honour inside a table. In CSV the break stays a real newline inside a quoted field, which every spreadsheet reads correctly.

A pipe character in a cell would otherwise start a new column, so it is escaped as \| in Markdown. CSV output goes through the same quoting rules as the CSV ⇌ JSON converter: fields containing the delimiter, a quote or a newline are wrapped in quotes and inner quotes are doubled.

Markdown or CSV?

Reach for Markdown when the table is going into a README, a pull request, documentation, a Notion or Obsidian page, or a prompt for a language model. Aligned columns (the default) make the source readable; switch Align columns off for a compact form that diffs better.

Reach for CSV when the table is data: something to open in Excel or Google Sheets, load with pandas, or feed into the CSV ⇌ JSON converter on its way to an API. The tab delimiter (TSV) is the safest choice when cells themselves contain commas.

Need to edit the table after converting? Paste the Markdown into the Markdown Table Generator for a spreadsheet-style grid. For a whole article rather than one table, the HTML to Markdown converter handles headings, lists and paragraphs.

What does not survive the trip

• Cell colours, widths, borders and fonts. Markdown tables carry text and alignment only; CSV carries text only.
• Nested tables, which are flattened into the text of the cell that contains them.
• Multiple header rows. A two-row header (common with a spanning group title) is kept as data rows; the first becomes the Markdown header.
• Column alignment is read from the align attribute or an inline text-align style on a cell. Alignment set in an external stylesheet is not visible in pasted HTML and is not applied.
• Content rendered by JavaScript after page load. Copy the table from the developer tools, not from “view source”, and you get the rendered rows.