Workflows·May 16, 2026·5 min read

Convert HTML Emails to Markdown for AI Summarization

Pasting a forwarded HTML email into ChatGPT yields a wall of inline styles and tracking pixels. Here's how to convert it to clean Markdown the model can actually summarize — in seconds.

Long-form newsletters — Stratechery, Platformer, vendor product updates — are mostly HTML wrapped around a small amount of actual content. When you paste the email into ChatGPT or Claude to ask for a summary, you're paying tokens for inline CSS, table layouts, tracking pixels, and the email client's own boilerplate.

A typical 2,000-word newsletter expands to 30,000+ tokens of HTML. Convert it to Markdown first and you're back to roughly 3,000 — a 90% reduction with zero loss of meaning.

The full workflow

In Gmail, open the email → ⋮ menu → 'Show original' (Outlook: 'View source'; Apple Mail: View → Message → Raw Source).
Copy the HTML body — the part between <body> and </body>.
Open Markdown Converter and paste.
Copy the cleaned Markdown straight into your AI chat.

Why Markdown beats raw HTML for AI

Roughly half the tokens for the same content
Headings, lists, and links survive intact
Models were trained on enormous Markdown corpora — they read it natively
Strips tracking pixels, inline styles, and email-client boilerplate automatically
Clean diff if you re-summarize after edits

What gets stripped (and why that's fine)

The converter removes inline CSS, conditional Outlook comments, MSO blocks, table-based layout scaffolding, and image tracking pixels. None of it carries meaning a language model needs. What survives: every heading, paragraph, link (with anchor text), bullet list, and image alt-text — exactly what you want for a summary.

If the email is still too long

Cleaned Markdown will fit any modern model's context window. But if you're forwarding a whole digest of newsletters, run the combined output through Context Compressor for a final 20–30% reduction. You'll fit days of newsletters into a single prompt and ask "what's actually changed this week?"

Sample prompt for newsletter summaries

You are summarizing a newsletter. Read the Markdown below and produce:

1. A 3-sentence TL;DR.
2. The 3 most actionable insights (with the link to each).
3. Anything that contradicts conventional wisdom in this space.

Newsletter:
[paste markdown here]

The model can now cite links because the Markdown preserved them. Try asking for the same thing on the raw HTML and the answer will hallucinate URLs.

Building a personal newsletter knowledge base

Convert each newsletter to Markdown and save it. Over a few months you'll have a searchable archive that an AI can query — "what did Stratechery say about Apple's services strategy in Q1?" — without you re-reading anything. The work is in the conversion; the archive compounds.

Tools mentioned

Frequently asked

Can I just paste the rendered email?

If you copy from your email client, you usually paste the rendered text — which is fine for short notes but loses links, structure, and context. HTML → Markdown preserves all three.

Where do I find the raw HTML of an email?

Gmail: open the message → three-dot menu → 'Show original'. Outlook: 'View source'. Apple Mail: View → Message → Raw Source.

Will tracking pixels affect the AI summary?

No, but they bloat the token count. The Markdown converter strips them automatically.

Keep reading

Workflows·Apr 12, 2026·6 min read