Substack Article to Markdown for AI: Reading Paid Newsletters with Claude (2026)

Substack is where the most considered writing on technology, economics, geopolitics, and culture happens in 2026. The best Substacks — Lenny's Newsletter, Stratechery (now on Substack), Doomberg, Slow Boring — have replaced what magazines used to do, for an order of magnitude lower subscription price.

Most of it is paid. The paywall is good — it filters who's there. But it makes feeding Substack content into Claude harder than it should be.

This post is the workflow that works.

Why your AI tools can't read paid Substacks

Standard server-side fetchers (ChatGPT browse, Claude WebFetch, Jina Reader, Firecrawl) all hit Substack URLs from their own servers without your authentication. Substack returns the paywall HTML:

<div class="paywall">
  <h2>This is a subscriber-only post.</h2>
  <p>Subscribe to read the rest...</p>
</div>

The AI then "summarizes" what's there — which is the paywall, not the article. Useless.

Even if you wanted to share your Substack auth cookies with these tools (you shouldn't), automation triggers anti-abuse detection. Substack's anti-bot system can lock the subscribing account.

The only category that works: read the article in your browser (where you're authenticated), extract the rendered content there.

What clean Substack Markdown looks like

After running through a Substack-aware extractor:

# What's actually changed in the tech labor market

**Author**: Lenny Rachitsky · **Published**: 2026-05-28
**Subscribers-only**: Yes (you have access)
**Source**: https://www.lennysnewsletter.com/p/tech-labor-market-2026

> Over the past 6 months, hiring patterns at tech companies have shifted in
> three ways that don't get talked about enough.

## The headline numbers everyone is missing

[Article body with proper paragraphs]

## What this means for individual contributors

[Continues...]

## What this means for managers

[Continues...]

## Footnotes

[1] Layoffs.fyi tracker — https://layoffs.fyi
[2] Pragmatic Engineer survey — https://newsletter.pragmaticengineer.com/

## Top comments

- **@reader1**: "Important caveat to point 2: the engineering manager numbers
  also include..."
- **@author_reply (Lenny)**: "Great point — I should have included that."
- **@reader2**: "The contrast with 2021 is striking..."

Token-efficient, properly structured, ready to paste into Claude as research input.

The workflow

Two paths:

Path 1: Web2MD Substack extractor (interactive)

Open the paid Substack article in Chrome (where you're already logged in as a subscriber). Click Web2MD. The Substack-specific extractor:

• Reads the rendered article DOM in your authenticated session
• Strips the Substack chrome (subscriber count badges, share buttons, related posts widget)
• Captures title, subtitle, author, publish date
• Preserves footnotes as a numbered references section
• Pulls top comments and author replies
• Outputs clean Markdown ready for Claude or your archive

End-to-end: ~5 seconds per article. The extension reads what's already on your screen — no cookie sharing, no third-party authentication.

Path 2: Browser-side bookmarklet for one-offs

If you don't want a full extension:

// Save as bookmarklet, click on a Substack article page
javascript:(function(){
  const title = document.querySelector('h1')?.innerText || '';
  const author = document.querySelector('.author-name')?.innerText || '';
  const body = document.querySelector('.post-body, .available-content')?.innerText || '';
  const md = `# ${title}\n\n**Author**: ${author}\n\n${body}`;
  navigator.clipboard.writeText(md);
  alert('Markdown copied to clipboard');
})();

Simpler than Web2MD but loses footnote handling, comment extraction, and doesn't generalize to other sites.

A real workflow: quarterly tech newsletter synthesis

Each quarter I do a "what did the best tech newsletters say?" synthesis. Sources:

• Lenny's Newsletter (PM/growth)
• Stratechery (tech strategy)
• Pragmatic Engineer (eng management)
• Doomberg (commodities, but useful tech context)
• Slow Boring (politics, edge cases for tech regulation)

5 newsletters × ~12 posts each over a quarter = ~60 posts. Web2MD queue + bulk export: ~40 minutes including skim-reading. Combined corpus: ~180k tokens.

Pasted into Claude Opus 4.7 with: "These are 60 newsletter posts from Q2 2026. Identify the 4 themes that multiple newsletters converge on, and the 2 where there's substantive disagreement. Quote specific posts and authors."

Output: a 6-page synthesis with quote-level citations. Total time: ~75 minutes for what's effectively a quarter-end industry recap.

What this is not

Honest about limits:

• Not a substitute for reading. The synthesis is useful for high-level pattern recognition. For load-bearing decisions, read the actual posts.
• Not for redistribution. Personal AI use of content you subscribe to is fine; sharing the Markdown with others violates your subscriber agreement.
• Not for free-tier abuse. If you cancel a Substack and use Web2MD to keep accessing old posts you saved while subscribed, that's the kind of edge case authors will rightfully push back on. The intent is "feed your subscriptions into your AI workflow," not "save everything before unsubscribing."
• Not for newsletter platforms other than Substack (without testing). Beehiiv, Ghost, Buttondown all have different structures. Web2MD handles each separately.

Pairing with other workflows

Substack content composes well with:

• Reddit-to-Claude: Substack for considered analysis, Reddit for user reactions
• Hacker News threads: HN for engineering perspectives on what Substacks discuss
• Fill 1M context window: 60 Substack posts is comfortably within Claude's context budget
• LinkedIn posts: LinkedIn for the same authors' shorter takes

The big picture

Paid Substacks are the modern equivalent of subscribing to The Economist or The Atlantic — a curated stream of considered analysis you pay for because the quality is worth it. The benefit of feeding them into Claude is the same benefit clipping articles into a notebook always was: you process more, you connect ideas across sources, you build a research base.

The mechanical step — extracting clean Markdown from a paywalled SPA — has been the missing piece. With a browser-side extractor, it's now a 5-second click.

Related

• Why Claude can't read Reddit (and how to fix it)
• How to fill Claude's 1M context window
• LinkedIn post to Markdown for AI summarization
• Hacker News thread to Markdown for Claude research
• Convert Substack to Markdown — supported sites page

Install

Web2MD on the Chrome Web Store →

Free tier: 3 conversions/day. Pro at $9/mo unlocks unlimited + queue + bulk export + dedicated Substack extractor that handles footnotes, author replies, and your subscription archive.