๐Ÿ“ฐ Newsletter format โ€” planned structure

Every issue follows the same structure so subscribers know what they're getting. AI writes each section from a brief. I pick the topic. That's the only human input per issue.

01
The Hook (Subject line)ChatGPT generates 5 subject line options. I pick the one with the most friction โ€” not the safest. AI ranks them by predicted open rate. Final call: mine.
~10 words
02
The Lead StoryOne AI development, explained clearly and without hype. Claude writes the first draft. ChatGPT does a "plain English" pass. The version that reads better goes in.
~200 words
03
3 Things Worth KnowingThree short bullets โ€” AI news, a prompt tip, and one tool recommendation. All AI-researched via Perplexity AI's current news mode. No dated information.
~150 words
04
The Hot TakeA bold opinion on an AI topic, written in first person. ChatGPT is prompted to take a strong position โ€” not hedge. This is the section most likely to drive replies and shares.
~100 words
05
One Prompt, Ready to UseA single copy-paste prompt for ChatGPT or Claude that solves a real problem. Tested before inclusion. The most-shared section of any prompt-focused newsletter.
~80 words
06
Sign-OffA consistent one-liner that closes every issue โ€” same format, same persona. Building consistency of voice across 30 AI-written issues is part of the hypothesis.
~20 words
"Newsletters win on consistency and voice. AI can do consistency. Whether it can fake voice well enough โ€” that's the whole experiment."

๐Ÿ“Š What I'm tracking

Subscriber metrics: new subscribers per day, source breakdown (organic search, social, word of mouth), unsubscribe rate. Engagement: open rate per issue, click rate, reply rate. Time cost: minutes to produce each issue โ€” if it takes 2 hours per issue it defeats the side hustle purpose. Weekly analytics published: every 7 days, raw numbers go in a public table on this page.

โ“ Frequently Asked Questions

Will you tell subscribers the newsletter is AI-written?

Yes โ€” the About page on Beehiiv will be transparent about the experiment. Each issue will have a small footer note: "This newsletter is part of an AI writing experiment โ€” read the full details here." The experiment's integrity depends on it. And honestly, readers deserve to know what they're engaging with.

Why daily? Isn't that too frequent for a newsletter?

Daily gives more data points across 30 days and tests whether AI can sustain daily output quality without degrading. Most human newsletters that are great go weekly โ€” daily is deliberately ambitious. If open rates collapse by Week 3, that's a useful signal. If they hold, that's an even more useful one. A weekly newsletter would produce only 4 data points in 30 days.

What's the niche?

AI tools and productivity โ€” same as this site. It's the niche with the most SEO overlap, the clearest audience, and the topic I can brief the AI most accurately on. A niche I don't understand would introduce hallucination risk that would make the experiment results less reliable.

How will you grow subscribers without paid ads?

Three channels: embedding the subscribe link in all new experiment articles published on this site (cross-audience), sharing each issue summary on Instagram (Exp 01 audience), and posting one LinkedIn post per week linking to the issue. No paid promotion. No cold outreach. Testing organic discoverability only.

What happens to the newsletter after the experiment?

If it reaches 100 subscribers with 35%+ open rates, I'll keep it going post-experiment โ€” potentially transitioning to a human-and-AI hybrid model where AI drafts and I edit more substantially. If it fails, the newsletter gets archived but all issues stay published. The experiment page stays up permanently.