Straight answer: AI hasn't made your marketing worse, it's made it identical to everyone else's, and readers can now smell it in about two seconds. The fix isn't to stop using AI, it's to use it for the scaffolding and then deliberately break the polish before you hit send.
The client email that made me stop and check my numbers
Back in early 2025 I was helping a UK coaching client rewrite her cold outreach sequence. She'd been sending the same five emails for two years, decent but tired, roughly a 7 to 8% reply rate on a warm-ish list of 300. We ran them through ChatGPT to tighten the language, fix the pacing, add a proper call to action. The new versions read beautifully. Clean structure, confident tone, no waffle.
We sent them to a fresh batch of 250 contacts the following month. Open rate barely moved, 41% versus her usual 38%. Reply rate dropped to 1.9%. Not a typo. From 7 to 8% down to under 2%, on a warmer list than the original one.
I've since seen the same pattern with three other clients and heard almost identical numbers from two people in my own network who run outbound for small agencies. Nobody talks about this because it makes the AI-productivity story messy, but the pattern is consistent enough now that I trust it.
What's happening when everyone uses the same tools
Here's the uncomfortable bit. When you and your three biggest competitors all run your emails through the same handful of AI tools with the same prompts ("write a friendly, professional cold email"), you all end up sounding like the same slightly-too-cheerful stranger. The sentence rhythm flattens out. The transitions all lean on the same three or four phrases. People have started recognising that rhythm the way they recognise a call centre script, and they stop reading the moment they hear it.
This isn't a hunch. Studies on inbox behaviour have shown for years that recipients decide whether an email is worth reading within the first line or two, and what they're scanning for, whether they realise it or not, is specificity. A name, a number, a detail that couldn't have been written about anyone else. AI-smoothed copy strips that out by default because it's trained to sound competent and inoffensive to the widest possible audience. Competent and inoffensive is exactly what gets deleted.
I wrote about this same problem from a different angle when I looked at how brands like Airbnb built a brand that wins on the back of real host stories rather than polished copy. The lesson holds for a single-person business writing outreach emails just as much as it holds for a billion-dollar platform. Specific and slightly rough beats smooth and generic every time someone else's inbox is involved.
Why the AI-marketing crowd won't tell you this
A lot of people selling "AI email prompts" and "AI content packs" right now have a real incentive to keep quiet about the reply-rate problem, because the whole pitch rests on speed. Write 50 emails in ten minutes instead of two hours, look how efficient you are. Nobody wants to follow that up with "and they'll convert at a quarter of the rate," because it kills the sale.
I've been doing this long enough, and rebuilt my own business publicly enough over the last five years, to know that the uncomfortable numbers are usually the useful ones. Efficiency that produces worse results isn't efficiency, it's just faster failure. If you're going to use AI in your marketing, and you should, the goal has to stay conversion and reply rate, not word count per minute.
The five-step fix I now use with every client
This isn't about abandoning AI tools. It's about where in the process they're allowed to touch the copy.
- Step 1: Brief it like you'd brief a junior copywriter, not a magic box. Give the AI your actual customer's actual words. I paste in three or four real quotes from calls or reviews before I ask for a first draft, never a generic "write me a cold email about X."
- Step 2: Let AI write the skeleton, not the voice. Ask for structure and logic only, in bullet points, then write the actual sentences yourself. This alone fixes most of the sameness problem because the words are yours even if the shape came from a tool.
- Step 3: Cut every sentence that could apply to any business in your industry. If a competitor could send the exact same line, delete it. Replace it with a real number, a real client name (with permission), or a specific detail about the recipient.
- Step 4: Put back one imperfection. A slightly blunt opener, a sentence fragment, a joke that's a bit too dry. Perfect grammar reads as manufactured. One rough edge signals a human wrote the final pass.
- Step 5: Read it out loud before sending. If it doesn't sound like something you'd say on the phone, it's still got AI fingerprints on it. Rewrite that bit.
For that same coaching client, we ran this process on the next batch of 250 contacts. Reply rate came back to 6.4%, still below her original best but well above the AI-only version, and open rate held steady at 40%. The fix took roughly 20 extra minutes per email over the pure-AI draft. Twenty minutes for a threefold jump in replies is not a bad trade.
Where AI earns its place in your marketing
None of this means throw the tools out. AI is brilliant at the parts of marketing that don't require a human fingerprint: summarising 40 customer interviews into three themes, generating ten subject line variants to A/B test in Mailchimp or a similar email platform, drafting the first pass of a landing page you'll edit heavily, or building the logic behind an interactive calculator that pulls people in with their own numbers rather than your marketing copy. Interactive tools like calculators sidestep the whole "does this sound like AI" problem entirely, because the visitor is reading their own inputs back, not your prose.
It's also good for the boring structural work. Batch-scheduling posts, tagging a content calendar, drafting variations for testing across channels the way strong brands do with their Instagram content strategy. Use it for volume and structure. Keep the final human pass for anything that lands in someone's inbox with your name on it and asks them to reply.
If you want a useful design analogy, look at how design tools like Figma built a brand around distinctiveness rather than sameness, even inside a category full of near-identical competitors. That's the same instinct you need in your outreach. Distinctiveness is the entire point, and AI's default setting works against it unless you actively push back.
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.
A quick way to test if your own content has already gone generic
Take your last five outbound emails or your last five LinkedIn posts. Cover the sign-off and ask yourself honestly whether a stranger reading them blind could tell they came from your business specifically, or whether they could have come from any consultant in your niche. If you can't tell the difference, neither can your prospects, and that's almost certainly why your reply rate has been quietly sliding for the last year without an obvious cause.
Alex Hormozi makes a version of this point constantly in his own copy, and it's one of the better business lessons worth pulling from his approach: specificity sells, vagueness doesn't, no matter how well it's punctuated. AI defaults to vague because vague is safe. Your job is to make it specific again after the draft is done.
If this is the bit of your marketing that's stuck, and you'd rather have someone look at the actual emails and numbers than guess, this is exactly the kind of gap an AI implementation coach should be closing with you, not just handing you another prompt list.
Frequently asked questions
Does AI-written marketing copy convert worse than human-written copy?
On its own, often yes, particularly for cold outreach and email where reply rate depends on the reader believing a real person wrote to them specifically. In my own testing with client campaigns, unedited AI drafts pulled replies down from around 8% to under 2%, and a heavy human edit pass brought it back to roughly 6%. The gap is real but fixable.
How can I tell if my emails sound too AI-generated?
Read your last few sent emails and check for generic openers, overly balanced sentence structure, and a total absence of specific names, numbers, or details unique to that recipient. If any competitor could have sent the exact same email, it's too generic, AI-written or not.
Should small businesses stop using AI for marketing altogether?
No. Use it for structure, research, subject line testing, and first drafts, then rewrite the final version in your own voice with real specifics added back in. The problem isn't the tool, it's sending the first draft without a human pass.
How long does it take to fix AI-flattened copy?
Roughly 15 to 25 minutes per email or post to strip out generic phrasing and add back specific detail, based on the client work described above. That's a small time cost for a return that, in real testing, more than doubled reply rates.
Want this done for you? See AI automation for coaches and consultants.
Related reading: Why Your AI Content Sounds Like Everyone Else's (And It's Not the Tool's Fault) and Why Your AI Customer Replies Are Quietly Losing You Repeat Customers.
