The first time I used AI to draft a follow-up email, I could tell within two sentences that a machine had written it. So could the person I was about to send it to.
It wasn't any one word that gave it away. It was the rhythm. Everything flowed a little too smoothly, the compliment was a little too generic, and the whole thing read like it could have been sent to anyone on my list. I deleted it and started over.
That happens to almost everyone the first time. The problem was never the tool. It was that I hadn't told it anything about the actual person, the actual conversation, or how I actually talk. Once I fixed that, the emails got better fast.
Quick Answer
AI can write a follow-up email that sounds like you, but only if you give it real context: the specific conversation, your history with that person, the tone you actually write in, and the one thing you want them to do next. Generic prompts produce generic emails. Specific prompts, fed with real detail, produce something you could have written yourself, just faster.
In this article, you'll learn:
- Why every AI email sounds the same by default
- The four things your prompt needs before AI can sound like you
- How to build a "voice profile" AI can reuse every time
- The words and phrases that instantly signal a robot wrote it
- How to edit AI's first draft in under a minute
Why Every AI Email Sounds the Same by Default
AI defaults to safe, neutral, corporate-sounding language when it doesn't have anything specific to work with. Ask it to "write a follow-up email" and it reaches for the same phrases it reaches for with everyone else who typed that exact same request. That's not a flaw in the tool. It's just what happens when there's nothing to differentiate one prompt from a thousand others.
That sameness has a real cost. According to HubSpot's research on sales email performance, personalized emails get a response rate of roughly 17%, compared to about 7% for generic, unpersonalized ones. That's more than double, and the gap comes entirely from whether the email feels like it was written for one specific person or blasted to a list.
The Four Things Your Prompt Needs Before AI Can Sound Like You
Generic AI output is almost always a prompt problem, not an AI problem. Before you ask AI to draft a follow-up, give it these four things:
The specific situation
What actually happened. Not "we spoke recently," but what was discussed, what question they asked, or what they said they needed to think about.
Your history with this person
Is this the first follow-up or the fourth? A returning client or a brand-new lead? The tone should shift depending on the answer, and AI has no way of knowing unless you tell it.
The tone you actually write in
Direct and brief, or warm and conversational? Tell it plainly, or better yet, give it an example of an email you've written before.
The one thing you want them to do next
A reply, a booked call, a yes or no. One clear ask, not three options buried in the last paragraph.
I wrote a fuller version of this framework in The Real Estate Agent's Guide to Writing AI Prompts That Actually Sound Like You, and the same logic is exactly why I stopped winging it with AI and built a prompt library instead. The output only gets better once the input does.
"AI didn't fail to sound like me. I failed to tell it who I am."Barton Eby
Build a "Voice Profile" AI Can Reuse Every Time
Re-explaining your tone in every single prompt gets old fast. The fix is building a short "voice profile" once and reusing it. A few sentences describing how you actually write: short sentences or long ones, how formal your greetings are, whether you use exclamation points, how you typically sign off.
Paste that voice profile into any follow-up prompt alongside the specific situation, and the output stops sounding like a template and starts sounding like something you'd actually send. It's a five-minute setup that pays off on every email after it.
The Words That Instantly Give Away a Robot Wrote It
Certain phrases have become dead giveaways. Anyone who reads more than a handful of sales emails a week has developed an eye for them, and once you notice the pattern, you can't unsee it.
Sounds Like AI
- "I hope this email finds you well"
- "I wanted to circle back and touch base"
- "In today's fast-paced environment"
- "I noticed your impressive momentum"
- Ends with three different options to choose from
Sounds Like You
- Opens with the actual thing you discussed
- References a specific detail from the conversation
- Uses the sentence length and tone you actually write in
- Says something true and specific, not a generic compliment
- Ends with one clear next step
None of this means avoiding AI. According to ZoomInfo's guide to AI sales email tools, the agents and reps getting real results are the ones who treat AI as a first-draft generator, not a finished product. The draft is the starting point. The edit is what makes it yours.
Editing AI's Draft in Under a Minute
Once you have a draft, three quick edits remove almost everything that reads as AI-written. Cut the throat-clearing opening line. Replace any vague compliment with one specific detail from the actual conversation. Make sure it ends with a single clear next step instead of several options.
That's it. Those three edits take less time than reading this paragraph, and they're the difference between an email that gets deleted and one that gets a reply. If follow-up in general is where you're losing deals rather than winning them, Most Real Estate Agents Follow Up Once goes deeper on why that happens and what to do about it. And if you want to see how this fits into everything else AI can take off an agent's plate, 5 Ways Real Estate Agents Are Using AI to Close More Deals Faster covers the other four.
Key Takeaways
- AI defaults to generic, safe language unless you give it something specific to work with
- Personalized emails get more than double the response rate of generic ones
- Every follow-up prompt needs four things: the situation, your history with the person, your actual tone, and one clear ask
- A reusable "voice profile" saves you from re-explaining your tone every single time
- Certain phrases are instant tells. Learn to spot and cut them
- Three quick edits, cutting the throat-clearing opener, adding a specific detail, and ending with one ask, turn a generic draft into something you'd actually send
This Is How You Get AI to Actually Sound Like You
The tool was never the problem. A blank prompt produces a blank-sounding email, no matter how good the AI is behind it. Give it the real situation, your real history with the person, your real tone, and one real ask, and the output changes completely.
That's not a workaround. It's just how prompting works. The more of yourself you put into the request, the more of yourself shows up in the answer.
Want help putting this into practice?
If you're already using AI but still not getting output you can use, the prompt framework is usually the missing piece. Reach out and let's talk about what you're working on →
