The short version: AI writing tools have made small business emails more polished and more forgettable at the same time, because thousands of businesses are running the same prompts and getting the same voice back. Open rates and reply rates drop even though the grammar improves. The fix isn't better prompts, it's putting back the specific, slightly rough, human detail that AI strips out.
The email that made me look twice
A client of mine, I'll call her Sarah because she'd rather I didn't use her real name, runs a boutique HR consultancy near Manchester with four staff and a client list of maybe 200 small businesses. In February 2025 she switched her entire newsletter over to an AI writing tool. Not lazily, either. She fed it her old emails, her tone of voice guide, brand adjectives, the lot. The copy that came back was, honestly, better than what she'd been writing herself. Cleaner sentences. Tighter subject lines. No typos.
Her open rate went from 31% to 19% in ten weeks.
She called me convinced Mailchimp had changed something on the backend, or that she'd landed on some spam blacklist. I checked her sender score, her domain authentication, her list hygiene. All fine. Nothing technical had broken. What had changed was that her emails had started sounding exactly like every other AI-polished newsletter landing in the same inbox, from her competitors, from the software companies she uses, from the podcast she listens to. Same rhythm. Same three-sentence opening hook. Same "here's the thing" transition. Same tidy call to action at the bottom.
Her subscribers weren't unsubscribing. They were just not opening, because their inbox had trained them that this shape of email doesn't need reading, it can be skimmed or binned in half a second.
The uncomfortable bit nobody selling AI tools wants to say out loud
Every AI writing tool is trained on roughly the same internet, using roughly the same handful of underlying models. That means when ten thousand small businesses ask for "a friendly, professional email about our new service," they get back ten thousand emails that are different in the details and identical in the skeleton. Same three-word punchy opener. Same rhetorical question. Same "let's be honest" or "here's the truth" pivot halfway down.
I read fifteen emails a day for client work and my own list. I can spot the AI ones now within two lines, and I don't think I'm special. I think most of your subscribers can too, even if they couldn't tell you why. Deliverability data from several email platforms in 2025 showed spam complaint rates rising fastest not on obviously spammy content but on this kind of "correct but generic" copy, because people report as spam anything that feels like noise, and generic reads as noise even when it's grammatically flawless.
That's the part the tool vendors don't put in the case study. AI didn't make your emails worse at the sentence level. It made your emails identical to everyone else's at the pattern level, and pattern recognition happens faster than we think.
What I told Sarah to do, step by step
We didn't ditch AI. That would have been throwing out a useful tool over a fixable habit. Here's the audit I ran on her last twelve newsletters, and it's the same one I'd suggest anyone runs before writing another word:
- Step 1: Read your last five sends back to back. Not skim them, read them aloud. If you can't tell which one was sent in October and which in December, your subscribers can't either.
- Step 2: Count the specific nouns. Real client names (with permission), real numbers, real dates, real prices. Sarah's AI drafts had zero specific nouns across five emails. Her best-performing email of the year, written before she switched tools, mentioned a client by first name, a £4,000 contract she'd nearly lost, and the exact phrase the client had used on the phone. That one had a 44% open rate.
- Step 3: Find the sentence that could only have been written by you. If every sentence in the draft could sit unedited in a competitor's email, you don't have a newsletter, you have a template with your logo on it.
- Step 4: Cut the AI-shaped transitions. "Here's the thing," "let's dive in," "the bottom line is." These phrases are fine occasionally. Used every single email, they become the tell.
- Step 5: Keep one rough sentence per email. Not a mistake, but something a polished writer wouldn't say. Sarah started adding one line at the top like "I nearly didn't send this one, it's a bit blunt" before a piece on a client mistake. Those emails now outperform her polished ones by roughly ten points on open rate.
By June, using AI for the first draft but running every single email through that five-step check before sending, Sarah's open rate was back to 34%, higher than where she started. The AI still does the heavy lifting on structure and speed. She just stopped letting it be the last voice in the room.
Where this shows up beyond email
The same pattern is hitting social captions, website copy, even proposals. I've seen two unrelated consultants send me nearly word-for-word LinkedIn posts about "the one mistake small businesses make with AI," three weeks apart, clearly both from the same prompt template doing the rounds. If you want to see what happens when a brand does the opposite, when every piece of content is unmistakably theirs, look at how deliberately specific GoPro built its marketing strategy around user footage nobody else could replicate, or how Instagram's own marketing strategy leaned on unmistakably specific creator stories rather than generic brand voice. Specificity is the actual moat now. Voice used to be free. It isn't anymore, because AI made bland voice free, which means distinct voice is the only kind worth paying attention to.
Joe Pulizzi built an entire content marketing career on one blunt idea I come back to constantly, which I wrote about in more detail in the business lessons I took from Joe Pulizzi: nobody has ever unsubscribed from content because it was too specific to the person writing it. They unsubscribe because it could have come from anyone.
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.
The fix isn't writing more, it's proving you're a real person faster
If you're time-poor, which every small business owner reading this is, the honest trade-off is this. Don't spend an extra two hours per email trying to sound human. Spend fifteen minutes adding three things AI can't invent because it wasn't there: a real number from this week, a real objection a real customer raised, and one opinion you'd only say out loud to a friend, not to a client. That's usually enough to break the pattern-match in a reader's brain.
If your emails have gone quiet and flat, it's rarely the subject line. I've swapped a dozen subject lines on client accounts and watched open rates move maybe two or three points. Changing the actual first two lines of the body, so the preview text stops sounding like a press release, moves it far more, sometimes eight to ten points in a single send.
And if you want subscribers to stop skimming altogether, give them something to do rather than something to read. Interactive content still cuts through in a flooded inbox precisely because AI can't fake interactivity the same way it fakes prose. I've written before about using interactive calculators to engage and convert visitors, and the same logic applies inside email: a short quiz, a calculator link, a genuine either-or question people have to answer, all outperform another well-written paragraph because they can't be skimmed past in half a second.
What I'm doing differently on my own list
I still use AI for first drafts, structure, and cutting waffle, because pretending otherwise would be dishonest and also a waste of a good tool. But I have a rule now: no email goes out that doesn't contain at least one detail from the actual week I've had, a real client conversation, a number from a real campaign, a thing that went wrong. It slows me down by maybe twenty minutes an email. My open rates on this list have held steady between 38% and 42% through 2025 while colleagues running fully AI-drafted newsletters told me theirs dropped into the low twenties. Same tools. Different habit at the finish line.
Frequently asked questions
Is AI causing lower email open rates, or is it something else?
It's rarely the sole cause, but it's a real factor. AI tools trained on similar data produce similarly structured copy across thousands of businesses, and readers unconsciously pattern-match generic-sounding emails as skippable. Combine that with normal list decay and seasonal drops, and it can look like a sudden crash when it's really a slow slide caused by sameness.
Should small businesses stop using AI for email marketing?
No. Use it for drafts, structure, subject line variations, and cutting your own waffle. The problem isn't the tool, it's sending the output unedited. Add specific numbers, real client details, and one honest opinion before you hit send, and the same tool becomes far more effective.
How do I know if my emails sound too generic?
Read your last five newsletters back to back and see if you can date them without checking the send date. If every email could have been written by a competitor, or by nobody in particular, that's the tell. Count how many specific nouns (real names, numbers, dates) appear. If the answer is close to zero, that's your answer too.
What's a quick fix if I don't have time to rewrite everything?
Change the first two lines of the email body, not the subject line. That's what shows in the preview text and it does more work than a clever subject line ever will. Add one real detail from your actual week and cut any sentence starting with "here's the thing" or "let's dive in."
Related reading: The AI Sameness Problem: Why Your Small Business Marketing Sounds Like Everyone Else's and AI Overviews Are Quietly Killing Your Website Traffic. Here's What To Build Instead.
