I Let AI Write My Newsletter for a Week. Here's What Happened.
Last month, in a fit of curiosity and exhaustion (I'd been up since 5 AM with a kid who decided that was a reasonable time to practice the recorder), I decided to run an experiment. I would let AI write my newsletter for a week. Not assist me. Not help me brainstorm. Write it. The whole thing. I'd just hit send.
I should tell you upfront: I had opinions about this before I started. Strong ones. I'm a writer. I've been a writer my entire professional life. The idea of outsourcing my writing to a machine felt like asking a robot to call my mom for me. Technically possible. Missing the point entirely.
But I wanted to be fair. I wanted to be open-minded. I wanted to see if maybe, just maybe, the tools had gotten good enough that my resistance was ego, not evidence.
So I tried it. Here's what happened.
Day one: pleasantly surprised
I gave the AI my usual newsletter topic (something about brand voice consistency) and some notes about the angle I wanted to take. What came back was... fine. Better than fine, actually. The structure was solid. The sentences were clean. There were even a few transitions that were genuinely smooth.
I was impressed. I was also a little threatened, which I'm honest enough to admit.
I sent it out. Nobody complained. Nobody replied with "this doesn't sound like you," which is what I was expecting. Open rates were normal. Everything was fine.
Day three: the uncanny valley
By day three, I started noticing something. The newsletters were competent, but they all had the same... texture. Like eating at a restaurant where everything is technically well-prepared but nothing has a point of view. The food is fine. You wouldn't complain. But you also wouldn't tell anyone about it.
The AI was writing correct sentences about content strategy, but it wasn't writing my sentences about content strategy. There were no cooking analogies. No self-deprecating asides. No stories about my kid or my grandmother or that time I accidentally sent a draft email to the entire company (we don't talk about that).
The voice was professional. But it wasn't a voice. It was the absence of voice, dressed up in competent prose.
AI can write correctly. It can't write vulnerably. And vulnerability is where voice lives.
Day five: the replies dried up
Here's the thing about my newsletter: people reply. Not a lot. Maybe ten or fifteen per issue. But they're real replies. People sharing their own stories. Disagreeing. Asking questions. Someone once sent me a photo of their grandmother's spice box after I wrote about brand voice and cooking. Those replies are my favorite part.
During AI week, I got two replies. Both were about a factual detail, not a personal connection. Nobody shared a story. Nobody sent a spice box photo.
The newsletter was still being read. But it wasn't being felt. And the difference between read and felt is the difference between content that exists and content that matters.
What I learned
I went back to writing my own newsletter on day eight. The first issue back, I opened with a story about the experiment itself. I was honest about my ego, my fear, and the weird sadness of reading competent prose that didn't sound like me. The replies came flooding back.
Here's what I took away from the experiment:
AI is excellent at structure. It can organize ideas clearly. It can create logical flow. It can write transitions that work. If I'm staring at a blank page and can't figure out how to arrange my thoughts, AI is genuinely useful as a structural tool.
AI is mediocre at voice. It can mimic a style if you give it enough examples. But mimicry isn't voice. Voice is the thing that happens when a specific human with specific experiences and specific insecurities sits down and types what they actually think. No model can replicate that because no model has lived your life.
AI can't be vulnerable. And vulnerability is what creates connection. The reason people reply to my newsletter is that I share things that make me a little uncomfortable. I admit when I'm wrong. I tell stories that don't make me look like a genius. AI can fake vulnerability, but readers can tell the difference. Always.
So where does AI fit?
I use AI now. Not to write my newsletter. But to brainstorm topics. To help me restructure a messy draft. To suggest angles I haven't considered. To write the boring-but-necessary stuff: meta descriptions, social posts, email subject line variations. The stuff that needs to be done but doesn't need to be me.
It's like having a very capable sous chef who can prep ingredients and organize the station but shouldn't be the one tasting the soup. The tasting is where the cook's judgment lives. The tasting is the part that can't be delegated.
Use AI for the parts that need to be correct. Use yourself for the parts that need to be human. Know the difference.
My kid still plays the recorder at 5 AM, by the way. No amount of artificial intelligence can solve that particular problem. Some things require purely human intervention, specifically earplugs and the patience of someone who remembers they chose to have children voluntarily.