Website text extractor: URL to Markdown or plain text

How to extract text from a website

1. Enter the page address. One page per run, such as a help article or a product page.
2. Choose Markdown or plain text. You can switch after the page is read.
3. Copy or download the result, then remove anything you do not need, such as a list of related links at the end.

How the extractor reads a page

The extractor fetches the page the way the Paperlink AI crawler does. It checks the site's robots.txt first and stops if our crawler is not allowed. It reads at most 2 MB of HTML, follows redirects only on the same host, and never opens private or local addresses.

From the HTML it keeps the main content: the <main> element, or <article>, or the body as a last resort. Scripts, styles and template code are removed. Then it converts the structure:

• Headings become # to ######.
• Lists become Markdown lists, numbered or not.
• Links keep their text. The link addresses are left out, so the text reads cleanly.
• Tables become Markdown tables.
• Code blocks stay as fenced code.

If a page has almost no text in its HTML, the extractor says so instead of returning an empty result.

Example

A help article comes out like this:

# Change your plan

You can change your plan at any time.

1. Open **Billing**.
2. Select **Change plan**.
3. Pick the new plan and confirm.

The new price applies from the next billing date.

What people use it for

• Preparing content for AI tools. Markdown gives a model the structure without the HTML noise, so answers and summaries stay on topic.
• Moving content between systems. Copy a help article from an old help desk into a new one, or into a docs repository.
• Checking what crawlers see. If the extracted text is missing a section you can see in the browser, that section is built with JavaScript, and many AI crawlers will miss it too.
• Editing and review. Paste plain text into a readability checker or a spell checker without menus and footers in the way.

Markdown or plain text?

Choose Markdown when the structure matters: when the text goes into a docs repository, a CMS that accepts Markdown, or an AI tool that answers from documents. Headings tell a reader, or a model, where one topic ends and the next begins.

Choose plain text for word counts, readability scores, translation tools and any field that shows Markdown symbols as they are. Plain text drops heading marks, bullets and bold markers but keeps the paragraphs.

Limits

• One page per run, up to 2 MB of HTML.
• Text built in the browser with JavaScript is not seen.
• Pages behind a login, a paywall or a cookie wall return the wall, not the content.
• Link addresses are not kept; copy them from the page if you need them.
• Images are kept as links only; their text is not read.
• The result is cached for one hour, so a page you just changed may show the old version.

Use extracted text responsibly

The extractor respects robots.txt, but robots.txt is not a license. Text on a public page still belongs to its author. Use it for your own content, for research, or for short quotes with a link to the source.

Next steps

To convert a document instead of a page, use PDF to Markdown. To give AI tools a curated list of your key pages, publish an llms.txt with the llms.txt generator. And if you want customers to get answers from these pages directly, a support chatbot can import the whole section and keep it up to date.