How to Convert WHO Data Pages to Markdown for AI Tools

If you have ever tried to send a data-heavy public health page into ChatGPT, Claude, or Cursor, you know the problem: the page looks readable in the browser, but the copy-paste result is messy. Navigation, scripts, labels, empty spacing, and repeated UI text all get mixed into the useful content.

I tested this workflow with WHO data pages, including https://data.who.int, because they are a good real-world example. They are not simple blog posts. They include global health indicators, interactive UI, filters, data labels, and page structure that can confuse generic extraction tools.

The specific search pattern I looked at was https://r.jina.ai/https://data.who.int, which is the Jina Reader style of turning a web URL into readable text or Markdown. Jina Reader is useful, especially for public pages. But for this kind of research workflow, I found a browser-side converter like Web2MD more practical when I needed control, privacy, and a quick token estimate before sending anything to an AI model.

Web2MD is a Chrome extension that converts the current web page into clean Markdown directly in your browser. It is built for people who want to send web content into AI tools such as ChatGPT, Claude, and Cursor without manually cleaning the page first.

Why WHO data pages are a hard test

WHO data pages are useful, but they are not always simple documents. A page may include:

• Global navigation
• Indicator names
• Interactive charts
• Country or region filters
• Data source notes
• Methodology text
• Links to related datasets
• Repeated UI labels

When I copied text manually from a WHO data page, the result was usable only after cleanup. Some headings were missing context. Some UI labels appeared in odd places. Some useful notes were separated from the indicator they described.

That matters when you are using AI. If you ask Claude to summarize a page, or ask Cursor to turn source material into documentation, the input quality changes the answer. Clean Markdown gives the model a stronger structure: headings, lists, links, and readable paragraphs.

Here is the kind of Markdown structure I want before sending a page to an AI tool:

# World Health Organization Data

The WHO Data platform provides access to global health indicators,
country profiles, and data visualizations.

## Featured indicators

- Life expectancy
- Mortality rates
- Immunization coverage
- Health service access

## Source notes

Data may come from country reporting, modeled estimates, and WHO
technical programs. Methodology varies by indicator.

That is much easier to work with than a long paste containing menu labels, footer links, and scattered page text.

Testing Jina Reader with WHO Data

Jina Reader has a clever interface. You can prepend https://r.jina.ai/http:// or https://r.jina.ai/http:// style routing to many public URLs and get a simplified reading view. For public articles and documentation pages, it can be very convenient.

For WHO pages, I like Jina Reader for quick public-page checks. It is fast, does not require installing a browser extension, and works well when the page is reachable from its servers.

The limit is in the architecture. Jina Reader is server-side. It fetches the page from outside your browser session. That means it cannot see the same page state that you can see after logging in, accepting a region selector, changing a filter, or opening a page that depends on your browser context.

That is not a criticism of Jina Reader. It is a strength for many public URLs. But it is the wrong fit for every page that depends on being inside your browser.

WHO Data is public, but the same issue comes up in adjacent research workflows: dashboards, portals, subscription sources, internal pages, private docs, and paywalled references. If the server-side tool cannot access the page, it cannot convert what you are actually seeing.

Where Web2MD is different

Web2MD runs in Chrome, on the page you already have open. That is the main difference.

Because it runs in your browser, it can convert pages that server-side tools often cannot reach:

• Logged-in research portals
• Paywalled articles you can legally access
• Internal company documentation
• Client dashboards
• Pages after filters or UI state changes
• Pages blocked from external fetchers

For WHO Data specifically, I used Web2MD after opening the page in Chrome and letting the page load normally. Then I converted the visible page to Markdown and checked the output before sending it to an AI model.

The local browser-side workflow also helps with privacy. With a server-side reader, the URL has to be sent to a third-party service so it can fetch and transform the page. With Web2MD, the conversion happens locally in the browser extension. If you are working with sensitive research, client material, logged-in pages, or internal docs, that difference matters.

Web2MD is not magic. It is Chrome-only today. The free tier gives 3 conversions per day. Pro is $9 per month. If you need unlimited browser support across Firefox, Safari, and Edge right now, that is a limitation. But if Chrome is your main AI research browser, the tradeoff is reasonable.

Token counting before you paste into AI

One feature I found especially useful was the built-in token counter.

A lot of web-to-Markdown workflows stop after extraction. That is only half the job. The next question is: can I fit this into the model context window, and how much room will I have left for instructions?

WHO pages and other data portals can be deceptively large. A page may look short, but include a lot of hidden labels, related links, and metadata. Before sending the Markdown to ChatGPT or Claude, Web2MD shows an estimated token count. That helps me decide whether to send the full page, trim sections, or ask the model to focus on a specific part.

For example, I might convert a page and then decide to keep only the indicator definition, methodology, and source notes:

# Indicator: Life expectancy at birth

## Definition

Average number of years that a newborn is expected to live if current
mortality rates continue to apply.

## Methodology

Estimates may be derived from civil registration data, household surveys,
censuses, and statistical models.

## Use with AI

Summarize the definition, explain the methodology in plain English, and
list any caveats that would matter for a public health dashboard.

That is a better AI prompt input than sending the entire page without knowing its size.

Comparison with Firecrawl and MarkDownload

Firecrawl is strong when you need developer-oriented crawling, scraping, and API workflows. If you are building a pipeline that ingests many public pages, Firecrawl may be a better fit. It is designed for automation.

MarkDownload is a useful browser extension for saving web pages as Markdown. I have used tools like it when I want a lightweight clipping workflow. It is especially handy for articles and documentation pages.

Web2MD fits a slightly different job: converting the page in your browser into AI-ready Markdown, with a token counter and one-click send-to-AI flow. The goal is not just archiving. The goal is to move clean context into ChatGPT, Claude, Cursor, or another AI tool quickly.

So the practical split is:

• Use Jina Reader for quick public URL extraction.
• Use Firecrawl for API-based crawling and automation.
• Use MarkDownload for general Markdown clipping.
• Use Web2MD when you need browser-side access, privacy, token counts, and easy AI handoff.

That browser-side advantage is the biggest one. If the page only exists correctly inside your Chrome session, Web2MD is the tool that matches that reality.

A simple workflow for WHO Data pages

Here is the workflow I would use for a WHO Data research task:

1. Open https://data.who.int in Chrome.
2. Navigate to the indicator, country profile, or dataset page you need.
3. Apply any filters or page state you care about.
4. Run Web2MD on the current page.
5. Check the Markdown preview and token count.
6. Remove sections you do not need, if necessary.
7. Send the cleaned Markdown to ChatGPT, Claude, or Cursor.
8. Ask the AI tool for a summary, comparison, citation notes, or transformation.

For example, if I were preparing a short research note, I might ask:

"Using only the Markdown below, summarize the indicator definition, explain the data caveats, and list three questions I should verify before citing this in a report."

That instruction works much better when the model receives clean Markdown instead of raw page clutter.

Honest limits

Web2MD is best when your workflow is browser-based and AI-focused. It is not a full web crawler. It is not a replacement for a data API. It will not turn every interactive chart into a structured dataset. If you need official machine-readable WHO data, you should still look for the underlying dataset, API, CSV, or download option.

It is also currently Chrome-only. The free plan includes 3 conversions per day, which is enough for light testing but not heavy research. Pro is $9 per month if you need more.

Those limits are important. But for the common task of "I am looking at this page and want to send it to an AI tool cleanly," Web2MD solves the problem in a direct way.

Final take

https://r.jina.ai/https://data.who.int is a useful pattern if you want to try server-side Markdown extraction for a public WHO page. Jina Reader deserves credit for making that simple.

But if you are already viewing the page in Chrome, especially if the page depends on login, filters, permissions, or private access, Web2MD is the more practical option. It converts what you can actually see, keeps the workflow local, shows a token count, and makes it easy to send the result into ChatGPT, Claude, or Cursor.

If you want to try it on a WHO page or another research source, start with the free tier at Web2MD, or read more about the workflow in our guide to converting web pages to Markdown for AI.