I Hired an AI to Write My Email Newsletter — Here's What Happened

I Hired an AI to Write My Email Newsletter — Here's What Happened

My email newsletter had 2,400 subscribers and a 38% open rate. I'd been writing it weekly for two years. It was time-consuming — each issue took 3–4 hours to write, format, and send.

In January I decided to try an experiment: hand the newsletter over to AI for two months. Not partially — completely. Every word drafted by AI, me only editing and approving. I wanted to know if AI could maintain a newsletter my subscribers liked, or whether they'd notice and disengage.

I tracked everything: open rates, click-through rates, unsubscribe rates, reply rates, and the occasional direct reader feedback. Here's what actually happened.

The Setup: How I Used AI for the Newsletter

Newsletter format: Weekly. Roughly 600–800 words. Structure: one main topic essay, 3–4 links with commentary, one "tool of the week" section, closing personal note.

AI tools used:

• ChatGPT-4o for essay drafts
• Anyword for subject line generation and scoring
• Grammarly for final editing pass

My role: Provide topic, provide any personal details I wanted included, review and lightly edit the draft, approve and send.

Time commitment target: Under 60 minutes per issue (vs. 3–4 hours previously).

Weeks 1–2: The Promising Start

The first two AI-drafted issues performed well. Open rates held at 37% and 39% — within normal variation of my historical average.

The main reason: I front-loaded the personal details. Before generating each issue, I gave ChatGPT a detailed brief:

"This week's topic is pricing psychology. My personal experience: I raised my rates by 40% last year and lost 2 clients but gained 3 better ones — net revenue up 22%. I want to open with that story. The tone should be direct and slightly irreverent — like I'm talking to a peer, not teaching a class. Include: [3 specific points I wanted to make], link to [specific article], close with: [specific thought I had]."

With that brief, the AI draft needed maybe 20% editing. The personal anecdote was in there. The tone was close. The structure was solid.

Week 1 stats: Open 37%, CTR 4.2%, unsubscribes 2 Week 2 stats: Open 39%, CTR 3.8%, unsubscribes 1

Weeks 3–4: The Warning Signs

In weeks 3 and 4, I got lazy with the brief. The topic was more technical (SEO changes), I didn't have a strong personal story to inject, and I gave ChatGPT a one-line brief: "Write a newsletter about recent SEO changes and what freelancers should do."

The output was technically accurate. It was also generic. I could feel it reading the draft. I edited it but couldn't fully rescue the voice — I'd given the AI nothing personal to work with.

Week 3 stats: Open 34%, CTR 2.9%, unsubscribes 5 Week 4 stats: Open 32%, CTR 2.7%, unsubscribes 8

The unsubscribe rate was the tell. Two weeks of above-average unsubscribes correlated directly with two weeks of under-briefed AI drafts.

One subscriber actually emailed: "Your newsletter felt different this week — less you, more generic blog post. Everything okay?"

That email told me everything I needed to know about what makes a newsletter feel real.

Weeks 5–8: The Corrected Approach

I returned to detailed briefs and reintroduced two elements I'd dropped:

1. At least one specific personal detail per issue — something that had actually happened to me that week, specific enough that the AI couldn't have invented it
2. At least one genuine opinion — not a balanced "there are pros and cons" take, but an actual position I hold

With these elements back in place, performance stabilized.

Weeks 5–8 average: Open 37.4%, CTR 3.9%

Subject Line Experiment: Anyword vs. My Own Intuition

The clearest win from the experiment was Anyword's subject line scoring. I generated 5–8 subject line variants per issue with Anyword and compared the top-predicted performer against what I would have chosen.

Results over 8 weeks:

Anyword won or tied in 7 of 8 weeks on open rate. The pattern: Anyword consistently chose more specific, slightly unconventional subject lines. My instinct was toward clear and direct; Anyword pushed toward intriguing and specific.

The Full 8-Week Results

Small performance decline. Significant time saving. The decline was concentrated in weeks 3–4 when my briefing approach was lazy. When I briefed well, metrics were comparable to my historical averages.

What AI Does Better Than Me in Email Writing

Subject line variants. I would naturally write 2–3 subject line options. With Anyword taking 2 minutes to generate 8 options with performance predictions, I always start from a better sample.

Structure and transitions. AI consistently produces cleaner structural flow between sections than my first drafts. I'm a good writer who produces lumpy first drafts. AI produces smooth, if sometimes generic, structural drafts.

Speed. 55 minutes vs. 3–4 hours is real. That 2.5–3 hours recovered per week compounds significantly.

What AI Cannot Do (That Matters for Newsletters)

Authentic personal voice. AI approximates your voice when given enough context. It never equals it. Readers of long-running newsletters have calibrated to very specific voice patterns — any deviation registers subconsciously as "off."

Specific situational knowledge. What happened in your niche this week. What your audience is talking about in private communities you're in. What you noticed in client work that surprised you. None of this can be generated — it can only be provided by you.

Genuine opinions. Newsletters that grow are built on perspective. Strong takes. Positions. AI defaults to balance. Your subscribers subscribe partly to hear what you think, not a balanced overview of the topic.

The Recommended AI Email Writing Workflow

Based on 8 weeks of testing, here's the approach that minimized performance decline while keeping the time savings:

1. Write a detailed brief (10 minutes) — topic, personal detail to include, specific opinion you hold, 2–3 points to cover, CTA
2. Generate draft with ChatGPT-4o (1 minute)
3. Generate 8 subject line variants with Anyword (2 minutes), pick the top scorer
4. Edit draft at sentence level (20 minutes) — add personal details, inject your opinion, cut generic AI phrases
5. Run through Grammarly (5 minutes) for final polish
6. Review and send (10 minutes)

Total: under 50 minutes. Quality: comparable to 3-hour manual drafts when briefed well.

Further Reading

• How to Use AI to Write Job Descriptions That Attract Top Talent
• Why I Switched from Grammarly to ProWritingAid (And Never Looked Back)
• How to Use AI to Write the Perfect Wedding Speech (Templates)
• Best Free AI Paraphrasing Tools That Avoid Detection (2026)
• Best AI Tools for Writing Poetry and Song Lyrics (2026)
• ChatGPT for YouTube Scripts: From Idea to Viral Video
• ChatGPT for Copywriting: AIDA, PAS, and Beyond
• The Art of Asking AI the Right Questions (And Getting Real Answers)