Is it Dishonest to Use AI?
How do I use AI and still reflect my own thinking? Plus a free and Honest Prompt Recipe that trains a personalized AI project to drafts emails that authentically sound like you.
Is it dishonest to use AI for work that’s supposed to reflect my own thinking?
The first time I used ChatGPT for real work, I immediately felt like I needed to look over my shoulder.
I was sitting at my kitchen table, coffee going cold, trying to write an email I’d already rewritten three times. It wasn’t even that important—just one of those messages where tone matters and you don’t want to sound defensive, vague, or like you’re unraveling before 10 a.m. I dumped my messy thoughts into ChatGPT, skimmed the response, and thought, damn… that’s really good.
Like, annoyingly good.
Not robotic. Not generic. Just clearer than what I’d written—like it skipped the part where I spiral, second-guess every sentence, and convince myself that suffering is part of being authentic. And that’s when something felt off. Not wrong, exactly. Just… easier than I thought it was allowed to be.
I remember thinking, “Okay, but if I send this, am I lying? Is this still my work? Or am I quietly outsourcing something I’m supposed to earn the hard way?”
Which is ridiculous, because thinking and writing are literally my job. I help other people do this all the time. And yet the moment a tool helped me do it with less friction, my brain went straight to: Cool, but at what cost to my integrity?
If you’ve ever used AI and felt that same mix of relief and guilt—like you found something genuinely helpful and immediately worried it said something unsettling about your character—you’re not imagining things.
Why this question feels heavier than it should
That uneasy feeling didn’t go away when I hit send.
Nothing bad happened. No one called me out. No AI police showed up at my door. The email landed just fine and the conversation moved on. But the question stuck with me longer than it should have: Why did that feel wrong when it actually helped?
The more I talked to people about AI, the more I realized this wasn’t just a “me” thing. The people who feel the most conflicted about using AI tend to be the ones who already care deeply about their work, their reputation, and their judgment. They’re not trying to cut corners. They’re trying to do things well—and then judging themselves for accepting help.
For most of us, “doing the work” has always meant effort you can see. Starting from a blank page. Holding everything in your head. Struggling a bit before things click. Somewhere along the way, we learned that effort equals authenticity—and ease equals suspicion.
AI breaks that story.
It gives you structure before confusion. Momentum before exhaustion. And for people who already feel scrutinized, that relief doesn’t feel neutral. It feels loaded. Like if something gets easier, maybe you didn’t earn it.
That’s why this turns into an ethics question so fast.
The bigger fear sitting underneath all of this
There’s also something else going on here, whether we admit it or not.
It’s not just, “Is this email still mine?”
It’s also, “Is this the beginning of the end?”
Every time a new technology shows up and actually works, we go through this same cycle. When the internet became mainstream, people were convinced it would rot our brains and unravel society. When social media took off, it was supposed to be the death of real human connection. (To be fair… that one might be partially accurate.)
Before that, it was television. Before that, the printing press. At one point, people genuinely worried that writing things down would destroy memory and thinking.
We’re very good at imagining technological apocalypse. Much better at imagining boring, incremental change.
AI hits closer to home because it touches thinking—the thing many of us have built our identity around. So when it helps, it doesn’t just feel useful. It feels threatening. Like it’s creeping into something personal. Like if you let it in too much, it might quietly take over.
That fear isn’t stupid. It’s human.
But it’s also worth separating the cinematic version of what’s happening from the reality most of us are actually living in: using a tool to write a clearer email, plan a week, or untangle a decision—and then immediately questioning our moral character because of it.
A lot of unnecessary anxiety lives in that gap.
What actually makes something dishonest (and what doesn’t)
Using AI isn’t what makes something dishonest.
Letting go of responsibility—and pretending you didn’t—is where things get murky.
When people ask whether it’s ethical to use AI, they’re usually asking something more personal: Is this still my thinking? Would I stand behind this? Could I explain how I got here without pointing at the tool?
Those are real questions. And they point to the actual line.
If you’re still responsible for the ideas, the decisions, and the outcome—if you’d stand behind the work if it were challenged—you haven’t crossed some invisible ethical boundary. You’ve just used help.
The problem isn’t AI. It’s outsourcing judgment and calling it your own.
But doesn’t this just make us lazy?
This is usually the next thought.
Not because people want to do less—but because they’re afraid of becoming someone who does.
We’ve been taught that effort is virtue. That if something feels hard, it must be meaningful. That struggle is proof you care. So when a tool removes friction, it can feel like it’s removing something else too—discipline, pride, worth.
But laziness isn’t using help. Laziness is disengaging from responsibility.
If you’re still deciding, still thinking, still accountable for the outcome, you’re not being lazy. You’re just not suffering for sport.
And for people who already carry more than their share of thinking, remembering, and planning? Offloading some of that isn’t moral failure. It’s maintenance.
A shift that makes AI feel normal again
Most people get into trouble with AI because they treat it like it’s there to finish things. To produce answers. To make decisions feel resolved.
A much healthier way to use it—and the one that doesn’t trip that internal alarm—is to treat it as thinking support.
Not: “Do this for me.”
But: “Help me think this through.”
That small shift changes how the tool behaves and how you feel about using it. And you can be explicit about that.
The Honest Prompt
Recipe: Train AI to Draft Emails That Sound Like You (Across Real Situations)
This is a small setup you do once, then reuse anytime you want help drafting emails without losing your voice.
You’re not teaching AI what to say.
You’re teaching it how you sound — including what you intentionally don’t do.
Step 1: Open a workspace for this project
Start a Project in ChatGPT, or a Project in Claude that you’ll keep using for email drafts. Name the project something that you will be able to remember. I named mine CPerozzi’s Personal Email Writer (zero points for creativity.)
Consider this your personal email workspace. Everything related to this project happens here.
Step 2: Train Your Email Writing Assistant to Sound Like You
Gather your AI training materials.
The first thing you need to do is give this project examples of how you write, what you sound like, how you handle problems, phrases you use, greetings, salutations, etc. So, before you paste anything into the tool, collect at least 5–7 real emails you’ve written and felt good about after sending. Not the polished ones, the honest ones that sound the most like you.
I try to include a mix of situations, for example you could include:
an email to your boss or someone senior
an email to a direct report
an email to a peer or collaborator
a casual email to a really good friend
an email to a new person or acquaintance.
an email where you clarified a misunderstanding
an email where you set a boundary or said no
These situations show how your voice adapts without changing who you are.
Combine Materials & Label Them
Copy and paste your emails into a doc that you can save somewhere. I usually paste my emails into a Google Doc. I have a folder called AI Training Docs and I named this document CPerozzi’s Personal Email Writer Training Materials (I know.) The reason I do this is so that if I ever need to build this tool again, I already have the information gathered. Make sure to add a simple label to each email so that the LLM understands the context, for example:
Email 1: To my boss
Email 2: To a direct report
Email 3: To a peer
You don’t need to explain the labels. The relationship is enough.
Upload or Paste the Training Materials
After you’ve copied and pasted your labeled emails, we need to add these to the LLM. You can do this in a couple of different ways.
I usually upload the document into the project as supporting materials. This way (and explain the benefit of doing it this way here)
You can also copy the entire document and then simply paste it into the LLM. When I do this, I tell the LLM what I’m giving it. I will write something like, “I’m building this project to be able to write emails as me. I’m about to give you several previous emails that I have written so you can start to glean the way I talk so you can sound like me. Do you understand? (I usually ask the LLM if it understands me so that there isn’t any confusion later on. ). Once you get affirmation that the AI understands you then paste the labeled emails into the tool.
Step 3: Write & Finesse the Training Prompt:
Now its time to tell the AI how to use the information that you’ve just given it. Copy and paste the following:
Training Prompt:
You’re looking at examples of how I naturally write professional emails. Your job is to learn my voice, not improve it.
Before writing anything new:
Analyze tone, pacing, and structure
Notice how I handle clarity, boundaries, and responsibility
Pay attention to what I don’t say as much as what I do
Notice how direct I am, and where I choose not to soften language
Notice how I open and close emails (or if I skip formalities altogether)
Do not:
Reuse phrases or sentences
Add enthusiasm I didn’t express
Make me sound more positive, warmer, or more agreeable
Introduce corporate filler or “polite noise”
It’s important to understand that my voice adapts to context and power dynamics, but it’s still one voice. Do not flatten tone across relationships. Do not invent warmth, deference, or authority I didn’t express.
Before drafting anything, describe the patterns you notice in how I write. Ask me questions if anything is unclear.
***End of Training Prompt***
Correct the AI:
Pause here and read what it reflects back to you. This step matters more than the drafting. Here is where you can have correct where the AI is reading it wrong. Take the opportunity to say “this is right” or “this is wrong” or “I actually usually say X.”
Step 4: Write the Execution Prompt to Write an Email that Sounds Like You
When you want help drafting an email, stay in or open the same project or conversation and ask something like this:
Prompt:
Help me draft an email about the topic I’m working on.
Here’s the context: who it’s going to, the relationship (boss, direct report, peer, external), the outcome I want, and any sensitivities or constraints.
Please match my natural voice and pacing. Keep the language plain and direct. Don’t over-polish or optimize. Don’t decide tone shifts for me. If something involves judgment, risk, or power dynamics, pause and ask instead of guessing.
Draft a first pass. I’ll edit before sending.
***End of Execution Prompt***
NOTE: Sometimes, I just copy and paste some or all of the email thread, if there is one, to give the tool more situational context. Sometimes, this might be the only thing the tool needs. But I usually explain the conversation. The more information you give the AI, the better result you get from AI.
Step 5: Teach it when you edit (this is the part most people skip)
Take a look at the draft and edit it as you think necessary. If you revise the draft and think, “This is closer, but still not quite me,” don’t just fix it and move on, instead, paste your edited version back in the tool and say:
This is how I revised that. Update your understanding of my voice based on this.
This teaches the model:
where you draw lines
what language you consistently remove
how you resolve ambiguity
Over time, this matters more than any single prompt.
Constantly doing this after every email draft will get your tool to the point where it will eventually be able to write an email draft that authentically sounds like you.
How to know this is working
If the final email still feels like something you’d stand behind — especially in moments of pressure — you’re not outsourcing your voice.
You’re just refusing to do everything the hard way.
Optional things for paid post (Need more of these ideas) And potentially add downloadable templates for the prompt here (and do I need to write just the prompts separately? Maybe this post should be two posts in 2 sections and then combine them for the newsletter? Please advise on how I should do this:
-Upload your favorite writers suggestions for communicative emails (Buy the book first).
-Add descriptions and thoughts on the people that you email often. You can give them personality quirks, a history of your relationship, information about great and/or dicey situations with them in the past.