Guide6 min read
Markdown vs plain text: check links and code
On this page5 sections
For a web article, start with Markdown when headings or source links need to survive. Choose plain text when you only need the words. If the article teaches HTML or includes literal markup, compare both exports with the source before choosing: in our example, Markdown kept the links but its preview changed the code examples into tables.
Compare both exports on the example
Open the arrow beside Copy, choose Copy as text or Copy as Markdown, and paste each into a plain-text editor before checking it in your destination app.
Markdown is itself plain text. The practical choice is between text with conventions such as ## Heading and [label](URL), and an export that removes those conventions. Changing a filename from .txt to .md does not restore missing links or heading levels.
Choose by what you need to keep
| Your next task | Choose and check |
|---|---|
| Notes or source-based AI research | Markdown; verify the source URLs. |
| Short quotation or email | Plain text; keep source attribution. |
| Literal code | Compare both exports with the source, then preview. |
| Spreadsheet data | A CSV or TSV table export; verify cells and types. |
One article, two different losses
We exported W3Schools’ HTML Tables article through imread on November 6, 2026 (19:47 UTC). We compared the text and markdown fields returned by the same extraction request, then rendered the Markdown with CommonMark 0.31.2 and raw HTML enabled. This was an export audit of one public article, not a clipboard test or a comparison of spreadsheet apps.
The article contains five example links with the same visible label: “Try it Yourself »”. In the Markdown export, all five labels kept their distinct example URLs. In the plain-text export, all five labels remained, but none of those five destination URLs appeared.
For example, one Markdown link was:
[Try it Yourself »](https://www.w3schools.com/html/tryit.asp?filename=tryhtml_table3)The corresponding plain-text label was:
Try it Yourself »This matters when you return to your notes or ask an assistant which example supports a claim. Five identical labels do not identify five different destinations. Keep the Markdown links, or add the relevant URLs alongside a plain-text excerpt yourself. Saving the original article URL also gives you a way back, but does not replace every linked destination.
| Observed in this export | Markdown | Plain text |
|---|---|---|
| The five example labels | 5 retained | 5 retained |
| Their distinct destination URLs | 5 retained | 0 present |
| Data table | Pipe-delimited | Tab-separated |
| Heading levels | Markers retained | Words only |
| Four HTML examples | Unfenced; rendered as tables | Literal tags |
1. Exact plain-text export
<table>
<tr>
<td>Emil</td>
<td>Tobias</td>
<td>Linus</td>
</tr>
</table>2. Exact Markdown source
<table>
<tr>
<td>Emil</td>
<td>Tobias</td>
<td>Linus</td>
</tr>
</table>3. Tested CommonMark preview (raw HTML enabled)
| Emil | Tobias | Linus |
Inspect the captured export fragments (JSON): all four HTML samples from both fields, the five Markdown links, capture time, and renderer settings. JSON escapes preserve the exported whitespace, including nonbreaking spaces. These fragments and our recorded counts let you inspect the examples; they do not independently reproduce the whole-export search. The file is not a complete API response or a clipboard sample.
The code result changes the recommendation. All four HTML examples were unfenced in the Markdown export. With raw HTML enabled, CommonMark 0.31.2 passed these table elements into its HTML output. In our Chromium preview, they appeared as tables rather than visible code. A separate source check found these examples in styled boxes rather than preformatted code elements; imread’s captured export left their tags unfenced and unescaped. Markdown can display literal tags inside code fences or with escaped brackets: this is a conversion failure on this input, not a limitation of Markdown itself. The text export retained the literal tags. We did not verify exact whitespace against the source, so this is not a claim that the text copy is ready to run unchanged.
CommonMark does not implement pipe tables; a GFM renderer or an app that disables raw HTML may behave differently. Neither format was an unqualified winner.
These observations describe this extraction, not every converter or every page. Plain text can contain URLs when they are written out, and a different exporter may append link destinations. Neither the table syntax nor the tab-separated output establishes how a spreadsheet will paste or interpret the data.
To check the tradeoff yourself, open the example above. In the reader, select the arrow beside Copy, then Copy as text or Copy as Markdown. The Copy button itself copies the format currently shown; it does not open the choices. Paste each copy into a plain-text editor first and inspect its literal characters and URLs. Then paste into your intended app and check what it retains or transforms. API output, clipboard content, and destination rendering are three separate checks. Reader Copy adds a title, and Markdown copy also adds the original source URL; our counts concern the API fields, not those additional lines. If the source changes, record the new result rather than expecting these counts.
What Markdown keeps—and what it does not promise
Markdown can represent headings, lists, links and code blocks as readable characters. A compatible app can render those characters as formatted content. The CommonMark specification defines core syntax; tables are an extension in GitHub Flavored Markdown, so support varies between destinations.
A conversion can still omit a section, drop a footnote or misread a table. An image URL or caption does not preserve the information inside a chart. Visual layout, interactive content and login-restricted material need separate checks against the original page.
For AI research, retained headings and links make the supplied material easier to inspect. They do not establish that the source is correct, that the extraction is complete, or that an assistant will follow a link. We did not measure answer accuracy or token savings in this example.
When plain text is enough
Use plain text for a short passage whose meaning does not depend on headings, links, a table or layout. It also avoids showing markup characters in a destination that does not interpret them. Keep the source URL with a quotation, and preserve any additional citation the passage needs.
Plain text is not automatically cleaner data. Removing formatting can hide a section boundary or turn several identical link labels into ambiguous text. Pick the simpler format only after checking that the information you need is still there.
A short check before you paste
- Compare the beginning, end and important passages with the original article.
- Pick one source link. Confirm that its destination is present and correct, not just its label.
- Check one heading and one list or table. Make sure their relationships remain understandable.
- For code or numbers, compare exact characters and values. Do not infer fidelity from appearance alone.
- Keep the original URL and the date you captured the article. Recheck the result in the app where you will use it.