How to Build a Newsletter Engine With Zero Writers Using AI
So a friend of mine launched an AI newsletter in March 2025 and hit 20,000 subscribers in eight months. He writes none of it. Zero. He runs an AI workflow that sources, summarizes, ranks, and drafts the daily edition, then he spends 15 minutes adding his voice, hits send, and watches the open rate. The newsletter is profitable on sponsorships. He works on it for an hour a day total. This is the workflow he runs, with the parts that are AI and the parts that aren’t honestly labeled. Because the dirty secret of “AI newsletters” is that the AI part is the easy bit; the voice, curation taste, and audience-building are the hard bits that still take a human.
The stack
- **Beehiiv** (or Substack, Ghost) — the publishing platform. ($0-$84/month at small scale)
- **Inoreader or Feedly** — RSS aggregation from your source list. ($60-$120/year)
- **Tavily or Brave Search** — for filling in coverage gaps the RSS misses. ($30-$100/month)
- **Claude or GPT API** — summarization and drafting. ($10-$50/month at moderate volume)
- **A workflow engine** — n8n, Zapier, or Make. ($20-$50/month)
- **A daily 15-30 minute human editorial pass.** Non-negotiable.
Total: $130-$300/month for the platform, plus 15-30 min of your time per edition.
The workflow
1. Build a tight source list. Newsletter quality is bounded by source quality. Curate 30-60 high-signal RSS feeds for your niche — primary sources, not aggregators. AI newsletters that pull from 500 sources end up bland; ones that pull from 30 curated sources have a voice.
2. Aggregate the day’s content in Inoreader. Set up a “today’s catch” view that surfaces everything from the source list in the last 24 hours. Maybe 200-400 items per day depending on your niche.
3. Score and filter with AI. Feed each item to a small LLM (Gemini Flash or GPT-5 Mini) with a prompt: “Rate this 1-10 based on relevance to [niche], novelty (not already covered widely), and reader interest. Output: score + one-sentence why.” Keep items above 7.
4. Cluster into stories. Many items will cover the same news. Use the LLM to cluster items that are about the same underlying story. Pick the best primary source per cluster.
5. Generate summaries. For each item, send the source URL content to Claude with a prompt that produces a 60-80 word summary in your newsletter voice. Save 4-6 summaries per edition; that’s the right density for daily AI newsletters.
6. Generate the rundown / TL;DR section. Most successful AI newsletters open with a punchy “what to read in 30 seconds” rundown. Claude generates this from the summaries. 100 words total, bullet style.
7. Add the human commentary. This is the 15-30 minute step nobody can skip. Read through the AI-generated edition. Edit for voice. Add the one piece of insight or opinion the AI can’t generate. Drop or replace any summary that’s wrong or boring. This is where your newsletter earns its voice and your reader’s loyalty.
8. Add the punchy intro and signoff. Top of newsletter: 60 words of voice that says hi and points to the most interesting thing. Bottom: a quick “what I’m thinking about” or “tool of the week” personal touch. AI generates the first draft of both. You rewrite.
9. Schedule and send via Beehiiv. Set a consistent send time. Track open rate, click rate, unsubscribes. Iterate.
10. Monetize when you have audience. Sponsors come when you cross ~5,000 engaged subscribers. Until then, focus on growth — the AI workflow gives you time to write Twitter threads and do collaborations, which is what actually grows newsletters.
What this saves
A traditional daily newsletter takes 2-4 hours per edition: source-gathering, reading, summarizing, writing the intro, formatting. This workflow cuts it to 15-30 minutes per edition while preserving the voice. Annual time saved: roughly 700-1,000 hours per year. For a solo operator running a newsletter as a business, this is the difference between a side project and a job replacement.
Gotchas
Voice is everything. AI-summarized newsletters without human voice read like RSS dumps and don’t grow. Spend the editorial time. Don’t try to skip it.
Source quality matters more than source quantity. A newsletter pulling from 30 great sources beats one pulling from 500 mediocre ones.
Fact-check the summaries. Especially for AI/tech news, the AI will occasionally claim things the source didn’t say. The 15-minute editorial pass catches most of this.
Beware “AI slop.” Multiple AI newsletters in the same niche reading the same sources can converge on suspiciously similar content. Differentiate with your specific takes, beat, or community.
FAQ
Do subscribers know I use AI?
Most don’t ask. Some will. Honesty is the right policy — “this newsletter is curated with AI and edited by me” doesn’t hurt growth and protects trust.
Will Beehiiv or Substack penalize AI content?
Not as of this writing, but the platforms care about engagement metrics (open rate, replies). If your readers stay, the platform stays neutral. If they unsubscribe en masse, platforms will downrank.
Can I run a paid newsletter with this workflow?
Yes — but the bar for paid is higher. Free curated newsletters work with AI summaries; paid newsletters need original analysis. Use AI for the curation and summarization; spend the saved time on original takes.
How fast does an AI newsletter grow?
Same speed as any newsletter: slowly at first, then compounding with cross-promotion, referral programs, and audience seeding. The AI workflow gives you time to do the growth work, not the growth itself.