Straight answer: AI drafting tools are trained on the same corpus of "helpful marketing tone," so if you let them write unedited, your emails start sounding like every other small business in your niche. The fix isn't ditching AI, it's building a short voice bank of your own real sentences and forcing every draft through it before it goes out. Took me six weeks of falling numbers to work this out.
What happened when I let AI write my newsletter
In early 2025 I got lazy. I had a list of just over 14,000 people, an open rate sitting around 38%, and a Tuesday deadline I kept missing. So I started feeding my rough notes into ChatGPT and letting it produce the full draft. Cleaner sentences. Faster turnaround. I thought I'd won.
Six weeks later my open rate had dropped to 24%. Click rate was worse, down by almost half. I assumed it was a deliverability issue, checked my sender score, checked my subject lines, blamed Gmail's Promotions tab. It wasn't any of that. I went back and read twelve of my own emails in a row and they all had the same rhythm: three-sentence intro, a bolded takeaway, "here's the thing," a bullet list, a soft call to action. Smooth. Forgettable. Nothing in there sounded like me telling a friend something over coffee.
Why this happens (and it's not your imagination)
Large language models predict the most statistically likely next sentence based on enormous amounts of marketing content that already exists online. That means the default output leans toward the tone that's already most common: friendly, competent, slightly bland. It's not that the AI is bad at writing, it's that it's brilliant at writing the average of everything, and the average of everything reads like nobody in particular.
This is the uncomfortable bit most people skip past: your subscribers can't always tell you exactly what changed, but they feel it, and they stop opening. Open rate is a trust metric before it's anything else. If your emails read like they could have come from any consultant in your industry, your reader's brain quietly reclassifies you as replaceable, and replaceable things get ignored.
The businesses that never fall into this trap
Look at brands that never sound like anyone else. Duolingo's marketing is unmistakable because it's built on one weird, specific voice choice (the unhinged owl) and they never soften it for safety. Calm writes slower, quieter copy than any competitor, on purpose, because it matches the product. Hotjar writes like a product team talking to other product people, full of specific UX language, not general "grow your business" fluff. None of that happens by accident, and none of it would survive being run through an AI draft and published unedited, because AI smooths out exactly the edges that make those brands recognisable.
Revolut is the same story in fintech: blunt, slightly cheeky, numbers-first copy that reads nothing like a traditional bank. That's a deliberate voice decision defended over years, not a prompt.
The 20-minute fix: build a voice bank
Here's exactly what I did to get my open rate back up to 41%, three points above where I started, within about five weeks.
- Step 1: Pull 20 real sentences you've written or said. Not polished ones. Pull them from old emails, voice notes you've transcribed, comments you left on a client call, even texts to a friend about work. I used a mix of old newsletters and the transcript of a podcast interview I did.
- Step 2: Paste those 20 sentences into a document titled "voice bank" and keep it open every time you draft with AI. Before you ask AI to write anything, tell it: "match the rhythm and specificity of these sentences, don't smooth them out."
- Step 3: Never let AI write your first line or your last line. Those two lines carry almost all the personality of an email. I write them myself, every time, in under two minutes, then let AI fill the middle if I'm short on time.
- Step 4: Delete every sentence that starts with "In today's fast-paced world," "As a small business owner," or "Here's the thing." These are the AI equivalent of throat-clearing. If you catch yourself writing them by hand too, cut them anyway.
- Step 5: Read the draft out loud before sending. If it doesn't sound like something you'd say to a client on a call, it needs another pass. This one step catches 80% of the "sounds like everyone else" problem.
A quick test to check if your own content has gone generic
Take your last three AI-assisted emails or social posts and cover the sign-off. Ask a colleague who doesn't know your business well to guess who wrote it from a list of five competitors. If they can't tell, you have a voice problem, not a content problem. I did this with my own team in March and two out of three people guessed wrong on emails I'd let AI draft unedited. Zero people guessed wrong on the ones I'd run through the voice bank process.
The same principle explains why interactive tools convert better than static text for a lot of businesses: specificity beats smoothness. When I wrote about interactive calculators for engaging site visitors, the whole appeal was that a calculator gives someone a specific, personal answer instead of generic advice. Your writing needs the same specificity. "Save time on marketing" is generic. "I cut my newsletter drafting time from 90 minutes to 20 and gained three points on open rate" is specific. AI will happily give you the generic version unless you push it toward the specific one.
Where this goes wrong for most small businesses
The pattern I see most with clients is this: someone adopts AI for content because it's fast, gets a short-term productivity win, and doesn't check the metrics that matter (opens, replies, actual sales calls booked) for two or three months. By the time they notice engagement has dipped, they've published forty pieces of slightly-too-smooth content and have no idea which one caused the slide. Check your open rate and reply rate weekly, not quarterly, for the first two months after you start using AI for any client-facing writing. Small, early drops are easy to fix. A quarter of flat content is a much bigger rebuild.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
There's a business lesson underneath all of this that's older than any AI tool. Dale Carnegie was writing about the power of using a person's name and speaking to their specific situation back in the 1930s, long before anyone had a content calendar. The principle hasn't changed: people respond to what feels specific to them, not to what sounds professionally competent. AI makes it easier than ever to produce professionally competent, forgettable content at scale. That's the trade-off nobody selling you an AI writing tool wants to say out loud.
What I do differently now
I still use AI for every single newsletter. It's faster, it catches typos I'd miss, and it's good at reorganising a messy voice note into readable paragraphs. What changed is that I never publish the first draft. Every email goes through the voice bank check, I rewrite the first and last line by hand, and I read it aloud once before it goes out. That's maybe eight extra minutes per email. It's the difference between 24% and 41% open rate on a list of 14,000 people, which on a typical email promoting a paid workshop is the difference between nine sign-ups and closer to twenty.
If you're running this for a whole team and want it built rather than patched together on your own, this is exactly the kind of thing an AI consultant for small business can set up in a single working session: a documented voice bank, house rules for what AI is and isn't allowed to write unedited, and a simple weekly check on the metrics that show you if it's working.
Frequently asked questions
Does AI-written content hurt SEO or email deliverability?
Not directly. Search engines and email providers don't penalise you for using AI. What hurts you is low engagement: if people stop clicking, stop replying, and stop opening, that behaviour signal is what damages your reach over time, not the fact that a tool helped draft the words.
How much should I edit an AI-written email before sending it?
At minimum, rewrite the first line and the last line yourself, and remove any generic opener like "In today's fast-paced world." That alone fixes most of the sameness problem. For anything going to your full list, read it aloud once before sending.
How long does it take to notice if AI content has flattened my brand voice?
Check your open rate and reply rate weekly for the first two months after introducing AI into your writing process. A drop of five or more points in that window is your early warning sign; leave it three months unchecked and you'll have a much bigger backlog of generic content to fix.
Can small businesses fix this without hiring anyone?
Yes, most of it is free. Build a voice bank of 20 real sentences you've written, feed it into your prompts, and never let AI write your opening or closing line unedited. That's the whole system, and it costs nothing but the twenty minutes it takes to set up.
Related reading: Apple Marketing Strategy: How They Built a Brand That Wins and ActiveCampaign Marketing Strategy: How They Built a Brand That Wins.
Related: ai content sameness small business.
