How to use the HTML to Markdown Converter

Converting a page with the HTML to Markdown converter takes about a minute. The steps below describe the tool generically, so small differences in labels or layout do not matter.

  1. Copy the HTML you want to convert. This can be a full page, the body of an article, or a single section. If you are working from a live page, use your browser’s “view source” or the Elements panel and copy the part that holds the actual content.
  2. Paste it into the input box at the top of the tool. Nothing is uploaded; the conversion runs in your browser, so you can paste drafts and unpublished pages safely.
  3. Run the conversion with the convert button. The Markdown appears in the output panel, usually within a fraction of a second even for long pages.
  4. Read the output. Check that headings are the right level, links still point where you expect, lists are nested correctly and code blocks are intact. Fix anything that looks off in the output or adjust the input and convert again.
  5. Copy the result to your clipboard or download it as a file. Save it with a .md extension if you plan to publish it as a Markdown page, or paste it into your llms-full.txt file.

TIP: Convert only the main content area. If you paste a whole page, you will also convert the header, navigation, cookie banner and footer, and then have to delete them by hand. Copy the element that wraps the article, usually <main> or <article>.

What HTML to Markdown conversion does

HTML describes a page for a browser. It carries structure (headings, lists, links) and presentation (classes, inline styles, wrapper elements, scripts). Markdown keeps only the structure, written in plain text with a handful of symbols: # for headings, - for bullets, [text](url) for links, backticks for code. A converter reads the HTML tree and rewrites each meaningful element as its Markdown equivalent, and drops everything that has no equivalent.

The mapping is straightforward for the common elements. Headings h1 through h6 become lines starting with one to six hash marks. Paragraphs become blocks of text separated by blank lines. strong and b become double asterisks, em and i become single asterisks or underscores. Anchors become inline links, images become image syntax with the alt text, and ul, ol and li become bulleted and numbered lists. pre and code become fenced or indented code blocks, and blockquote becomes lines prefixed with a greater-than sign.

What gets dropped is just as important. Scripts, styles, tracking pixels, hidden elements, wrapper div and span tags, class names and inline styles carry no meaning in Markdown, so they disappear. What remains is the content and its outline, which is exactly what you want when the reader is a language model or a person skimming a plain-text file.

Markdown flavors and why they matter

There is no single Markdown. The original specification has no tables, no strikethrough and no task lists. GitHub Flavored Markdown (GFM) adds those, and most modern tools understand GFM. CommonMark tightened the ambiguities in the original. When you convert HTML that contains a table, the converter has to choose: write a GFM pipe table, or fall back to plain text or leave the raw HTML in place. Check how your tool treats tables, because that is where most conversions differ.

If your Markdown will be read mostly by AI systems, flavor matters less than you might think. Language models have seen enormous amounts of Markdown, including GFM tables, and read both forms well. If it will be rendered by a static site generator or a CMS, match the flavor that system expects.

AI engines and answer engines read the web in two broad ways. Search-connected assistants fetch pages at answer time, strip them down to text and pass the text to a model. Training and indexing crawlers collect pages in bulk and process them offline. In both cases, the model never sees your design. It sees text, and how much of that text is useful depends on how cleanly it can be extracted from the markup.

A typical modern page is mostly not content. Navigation menus, mega-menus, cookie notices, newsletter popups, related-post widgets, footers and inline scripts can outweigh the article several times over. Extraction software tries to find the main content, and it usually does a reasonable job, but it can fail on unusual layouts, content rendered by JavaScript, or pages where the article is split into tabs and accordions. Markdown sidesteps the problem by handing the reader only the content.

There are three practical places where converted Markdown helps.

Markdown versions of your pages

Some sites publish a .md twin of each page, so that /guide/ also has a plain-text version at /guide.md. The idea is that an AI agent that wants a clean copy can ask for one. This is an emerging convention, not a standard that every engine follows, but it is cheap to offer and harmless to human visitors. The free aeoora WordPress plugin can generate Markdown for every page automatically, so you only need the converter for content that lives outside WordPress or for one-off conversions.

llms-full.txt

The proposed llms.txt convention describes a short Markdown index of your most important pages. A companion file, llms-full.txt, goes further and includes the full text of those pages in one document. Building that file by hand means converting each page to Markdown and concatenating them under clear headings. This is the task the converter is built for. Be honest with yourself about the benefit: llms.txt is a proposal, and AI engines rarely fetch it today. Treat it as low-cost preparation and a convenience for developers and agents who do look for it, not as a ranking lever. Our llms.txt generator covers the index file.

Feeding content to models and pipelines

The most reliable use of Markdown today is one you control: pasting content into a chat assistant, loading it into a retrieval pipeline, or giving it to an agent as context. Markdown is compact, so it costs fewer tokens than the same content as HTML. A 2,000-word article can easily be several times larger as raw HTML than as Markdown, and every wasted token is context window you cannot use for the actual task. If you want to see the difference on your own content, run both versions through the word and token counter.

What the converter keeps, changes and drops

Knowing what to expect saves you from surprises. These are the typical behaviors; your output may differ slightly depending on the options the tool exposes.

Kept

  • Heading hierarchy, written as hash marks.
  • Paragraph text, with bold and italic emphasis.
  • Links with their visible text and URL.
  • Ordered and unordered lists, including nested lists.
  • Images, as alt text plus the image URL.
  • Code and preformatted blocks.
  • Block quotes.
  • Tables, either as pipe tables or as plain text rows, depending on the tool.

Changed

  • Entities such as &amp; and &nbsp; become the plain characters they stand for.
  • Underlined or highlighted text loses that styling, because Markdown has no equivalent.
  • Line breaks inside paragraphs may collapse into spaces, since HTML ignores source whitespace.
  • Definition lists, details and summary elements, and figures with captions are flattened into paragraphs and lists.

Dropped

  • script, style and noscript content.
  • Class names, ids, inline styles and data attributes.
  • Layout wrappers such as div, section and span, though their text content is kept.
  • Forms, buttons and form controls, which have no useful text-only form.
  • HTML comments and tracking pixels.

Best practices

A converter gives you a first draft of the Markdown. The practices below turn that draft into a file that reads well and holds up over time.

Start from the content, not the whole page

Choose the smallest element that contains what you want to keep. On most themes that is <main>, <article> or the post content container. Starting there avoids converting menus, sidebars and footers, which are the main source of junk in converted pages. If you must convert a full page, plan a cleanup pass to delete navigation and footer text.

Keep exactly one H1

A Markdown page should start with a single top-level heading that names the page, then use second-level headings for sections and third-level headings for sub-points. Many themes put the logo in an h1 or use several h1 elements; the converted output will inherit that. Fix the outline by hand. A clean heading structure helps readers and helps retrieval systems that split documents into chunks at headings.

HTML often uses relative links such as /pricing/ or ../docs/. In a standalone Markdown file those links point nowhere, because the file has no site context. Rewrite them as full URLs, for example https://example.com/pricing/. This matters most for llms-full.txt, which is read as a single file away from your site. Do the same for image URLs.

Preserve useful alt text, drop decoration

Images become ![alt](url). Good alt text carries information a text-only reader would otherwise miss, such as what a chart shows. Decorative images with empty alt text are better deleted from the Markdown entirely, since they add a line of noise with no meaning.

Review tables and code

Tables and code blocks are the elements most likely to break. Read them in the output. If a complex table with merged cells was flattened badly, rewrite it as a short list or a simpler table. For code, make sure the block is fenced with backticks and consider adding a language label after the opening fence, such as html or js, so renderers and models know what they are looking at.

Remove what a reader does not need

Cookie text, “share this post” prompts, “related articles” lists, author boxes repeated on every page and promotional banners all survive conversion if they were in the HTML text. Delete them. The goal is a document that answers a question, not a screenshot of your page in text form.

Keep the source of truth in one place

Converted Markdown is a copy. If you edit the page, the copy goes stale. For anything beyond a one-off, automate the conversion at publish time instead of maintaining Markdown files by hand. The free aeoora WordPress plugin does this for WordPress sites: it serves Markdown for every page from the same content, and builds llms.txt and llms-full.txt from it, so nothing drifts.

Examples

Seeing a before and after makes the mapping concrete. Here is a small piece of HTML with typical clutter.

<main>
  <article class="post">
    <h1 class="entry-title">How to choose a running shoe</h1>
    <p class="lead">The right shoe depends on <strong>your gait</strong>, your weekly mileage and the surface you run on.</p>
    <h2 id="fit">Fit comes first</h2>
    <ul class="checklist">
      <li>Leave a thumb's width in front of your longest toe.</li>
      <li>Try shoes in the afternoon, when feet are slightly swollen.</li>
    </ul>
    <p>Read our <a href="/guides/lacing/">lacing guide</a> for fixes.</p>
  </article>
</main>

The converter reads the tree, ignores the classes and wrappers, and returns Markdown like this.

# How to choose a running shoe

The right shoe depends on **your gait**, your weekly mileage and the surface you run on.

## Fit comes first

- Leave a thumb's width in front of your longest toe.
- Try shoes in the afternoon, when feet are slightly swollen.

Read our [lacing guide](/guides/lacing/) for fixes.

Notice the relative link /guides/lacing/. For a standalone file, change it to the full address, https://example.com/guides/lacing/. Everything else is ready to use.

A table example

Tables show why you should always read the output. This HTML table has a header row and two data rows.

<table>
  <thead><tr><th>Plan</th><th>Seats</th></tr></thead>
  <tbody>
    <tr><td>Starter</td><td>3</td></tr>
    <tr><td>Team</td><td>10</td></tr>
  </tbody>
</table>

A converter that supports GitHub Flavored Markdown produces a pipe table.

| Plan | Seats |
| --- | --- |
| Starter | 3 |
| Team | 10 |

A converter that sticks to core Markdown may leave the raw table in place or flatten it to lines of text. Either result is readable to a language model, but only the pipe table renders as a table in most Markdown viewers.

An llms-full.txt layout

When you assemble llms-full.txt, keep one clear heading per page and note the canonical URL so a reader can trace each section back to its source.

# Example Store

> Family-run shop selling trail running gear, with free fitting advice.

## How to choose a running shoe

Source: https://example.com/guides/running-shoe/

The right shoe depends on **your gait**, your weekly mileage and the surface you run on.

## Returns policy

Source: https://example.com/returns/

You can return unworn items within 30 days.

The file is nothing more than converted pages joined under headings. There is no special syntax beyond Markdown itself.

Common mistakes

  • Converting the whole page. Menus, banners and footers end up in your Markdown and then in your llms-full.txt. Start from the main content element.
  • Leaving relative links. A link to /pricing/ is meaningless in a file that lives away from your domain. Convert every internal link and image path to a full URL.
  • Trusting the output without reading it. Nested lists, tables and code blocks are where conversions go wrong. A thirty-second scan catches broken structure before it spreads.
  • Keeping several H1 headings. Themes love to add extra top-level headings. A document with five H1s has no clear outline. Keep one and demote the rest.
  • Converting content that JavaScript builds. If you copy the source HTML of a client-rendered page, the content may not be in it at all. Copy the rendered HTML from the Elements panel instead, or use the page’s real text.
  • Pasting sensitive HTML into online tools. This converter runs in your browser, but be careful with any converter that sends your input to a server. Do not paste private pages, tokens or customer data into tools you have not checked.
  • Treating llms-full.txt as a ranking tactic. No major AI engine has confirmed that it relies on these files. Publish them because they are cheap and useful, not because you expect a jump in citations.
  • Maintaining Markdown copies by hand. Hand-edited copies drift from the live page within weeks. Generate them from the source whenever you can.
  • Stripping too much. Aggressive cleanup can remove the caveats, dates and qualifiers that make an answer correct. Cut the clutter, not the substance.

When to convert and when to write Markdown directly

Conversion is a bridge for existing content. If you are starting from scratch, writing in Markdown from the beginning produces cleaner output than any converter, because you control every heading and list. Many teams draft in Markdown and publish to HTML, which keeps the Markdown copy as the source of truth.

Use conversion when the HTML already exists and you need a plain version: migrating a site to a Markdown-based generator, creating a snapshot for an AI assistant, assembling llms-full.txt, archiving a set of pages, or building a knowledge base from a help center. In each case the converter does the mechanical 90 percent, and you do the final review.

How AI systems actually use clean text

It helps to be precise about what Markdown does and does not do for AI visibility. It does not make a page more authoritative. It does not change how a search index ranks you. What it does is lower the cost and the error rate of reading your content. A language model working from clean text is less likely to mix a footer link into an answer, mistake a menu label for a heading or run out of room before it reaches the part of the page that answers the question.

Retrieval systems typically split documents into chunks before embedding them, and they often split at headings. Markdown headings give those systems clear boundaries. A page whose sections each answer one question, under a heading phrased like that question, produces chunks that can be retrieved and quoted on their own. That is why the conversion step is also a good moment to review your headings: if a section heading says “Overview” or “More information”, rewrite it into something that stands on its own.

Citations work the same way. When an engine quotes you, it usually pulls a sentence or short passage that directly answers the prompt. Short, self-contained paragraphs convert cleanly and quote cleanly. Long paragraphs that depend on three preceding ones do not. If your converted Markdown reads awkwardly, the HTML source probably had the same problem, and fixing the writing helps both versions. The AI answer readiness checker scores that quality directly.

Privacy and accuracy notes

The converter processes your input locally in your browser. aeoora does not receive the HTML you paste. Even so, treat the output as a draft: automated conversion is deterministic but not perfect, and edge cases such as malformed HTML, deeply nested tables, inline SVG and unusual character encodings can produce odd results.

Two accuracy caveats are worth knowing. First, HTML that is not well formed is parsed by browsers using error-recovery rules, and different parsers recover differently, so unclosed tags may produce different structures in different tools. Second, Markdown cannot represent everything HTML can, so some information is lost by design: colors, columns, interactive widgets, embedded video, and the visual relationship between elements. If that information matters, keep a link to the original page next to the Markdown.

Frequently asked questions

Is my HTML sent to a server?

No. The conversion runs in your browser using JavaScript, so what you paste stays on your device. You can paste drafts and private pages without them leaving your machine. As with any tool, avoid pasting passwords, API keys or personal data you would not want in your clipboard history.

Will converting my pages to Markdown improve my AI visibility?

Not directly, and nobody should promise that it will. Markdown makes your content cheaper and cleaner for a model to read when it does read it, and it is useful for files like llms-full.txt and for your own AI workflows. Visibility in AI answers still depends mainly on whether your content answers real questions, whether it is trustworthy and whether other sources mention and cite you. Use Markdown as preparation, not as a shortcut.

Do AI crawlers fetch .md pages or llms-full.txt?

Some agents and developer tools do, but most major AI engines do not publicly rely on them. The convention is proposed rather than adopted as a standard. Because publishing the files costs little and has no downside for human visitors, many sites do it anyway. If you want evidence for your own site, check your server logs or an AI crawler dashboard to see whether any bot requests those URLs. The aeoora WordPress plugin includes a crawl log, and aeoora Cloud analyzes verified AI bot visits from your logs.

What happens to images?

They become Markdown image syntax containing the alt text and the image address. The image itself is not embedded, so a reader that only sees text gets the alt text and a link. Write descriptive alt text on the source page, and make sure the image URL is absolute if the Markdown will be used away from your domain.

Why does my table look wrong in the output?

Core Markdown has no table syntax. Converters either emit a GitHub Flavored Markdown pipe table or fall back to something simpler. Tables with merged cells, nested tables and tables used for layout do not translate well to pipe tables, which assume a simple grid with one header row. Rewrite complex tables as a simpler table or as a list.

Can I convert Markdown back to HTML with this tool?

This tool goes one way, from HTML to Markdown. To go back, use any Markdown renderer or your CMS. Keep in mind that a round trip is not lossless: classes, inline styles and wrapper elements are gone after the first conversion.

How large a page can I convert?

Browsers handle large inputs well, so a long article or a handful of pages pastes without trouble. For very large inputs, such as an entire site export, convert in batches or automate it with a script or a plugin that does the conversion server-side, rather than pasting everything into a text box.

Should I convert my whole site?

Usually not. Pick the pages that matter most to your business: pricing, product or service pages, your best explainers, policies and support answers. A focused set is easier to keep accurate than a full dump. If you use WordPress, the free plugin handles per-page Markdown for you, and you can limit llms-full.txt to the pages you choose.

Go further with aeoora

If your site runs on WordPress, you do not need to convert anything by hand. The free aeoora WordPress plugin serves a Markdown version of every page, builds llms.txt and llms-full.txt, manages robots rules for AI crawlers and adds schema without duplicates. It needs no account and has no paid tier. The WordPress plugin documentation walks through the setup.

To see whether any of this affects how AI engines treat your brand, aeoora Cloud tracks your visibility, citations and AI referral traffic across engines. See plans and pricing or start a 2-day free trial; a card is required and you can cancel before day 3. Our GEO tools documentation covers the llms.txt, schema and FAQ generators on paid plans, and you can find the rest of the free utilities on the free tools page.