ChatGPT Projects for Email: Write Every Message in Your Voice

Email is where generic AI voice is most obvious, and most damaging.

A mediocre blog post costs you credibility with readers you might never have engaged anyway. A generic client email costs you a relationship you're actively managing. When your email reads like it was written by the same AI everyone uses, clients notice. Colleagues notice. Prospects notice.

This isn't a problem with ChatGPT's capabilities. It's a configuration problem. ChatGPT defaults to a neutral professional voice because that's what fits most requests from most users. Without explicit instructions telling it your voice, it writes a version of professional that belongs to everyone and no one.

The fix — a dedicated Email Project with your voice properly loaded — takes about 30 minutes to set up and changes how every email you write from ChatGPT sounds.

---

Why Email Needs Its Own Project

If you're already using ChatGPT Projects for your writing, you might wonder whether you need a separate project just for email.

You probably do, and here's why: your email voice isn't the same as your writing voice. (OpenAI's prompt engineering guide covers the general principles; the voice-specific part is what this post is about.)

Your blog posts might be expansive, structured, teaching-focused. Your emails are compressed, direct, relationship-aware. You shift register constantly: formal with clients, casual with your team, warm-but-professional with cold contacts you want to win over. A project built for blog writing will produce blog-length email. A project built for formal writing will make your team emails feel stiff.

Email is also the highest-volume writing most professionals do. If you send 30 emails a day and spend 15 minutes per email thinking about tone and phrasing, that's 7.5 hours a week. With a properly configured Email Project, ChatGPT handles the drafting and you spend that time editing for nuance instead.

The separate project isn't overhead — it's the thing that makes the tool actually work for this format.

---

Step 1: Create Your Email Project

In ChatGPT, create a new Project and name it something like "Email / [Your Name]", specific enough that you'll reach for the right project when you're about to write.

In the project instructions, you're going to write two things:

1. Your baseline email voice
2. Your per-context register rules

Don't start with a blank page. Answer these questions first:

Baseline voice:

• How formal are you by default? (Scale from "hey" to "Dear [full name]")
• Do you use contractions? ("it's" vs. "it is")
• How long are your typical emails? (3 sentences? 3 paragraphs?)
• Do you use bullet points or do you write in prose?
• What's your signature opener? What's your closer?
• What phrases do you overuse that you're trying to eliminate?
• What corporate-speak do you avoid?

Per-context rules:

• Clients: how formal, how direct, how much explanation do you provide?
• Internal team: how casual, what's your shorthand, do you skip pleasantries?
• External partners (vendors, collaborators, press): somewhere between client-formal and team-casual?
• Cold outreach: how do you open, how long do you go, what's your call to action?

Once you have answers, turn them into explicit rules in your project instructions.

---

Step 2: Write Your Project Instructions

Here's the structure that works. Replace the examples with your actual voice:

EMAIL VOICE: [Your Name]

BASELINE
- Register: professional-casual. Formal enough for business, relaxed enough to sound human.
- Contractions: yes (I'm, it's, we're, don't; not "I am" "it is")
- Length: brief by default. If it takes more than 3 short paragraphs, I usually need to reconsider what I'm asking or rethink the format.
- Structure: prose unless there are 3+ items that benefit from a list. No forced bullets.

SIGNATURE PHRASES I USE
- Opener: "Quick note on [topic]" for short asks / "Wanted to loop you in on [topic]" for updates
- Closer: "Let me know if this lands well" / "Happy to jump on a call if easier"
- Subject lines: specific and short; never more than 7 words

PHRASES I AVOID
- "I hope this email finds you well"
- "Please don't hesitate to reach out"
- "As per my last email"
- "Kindly" (anything)
- "Circle back" (vague follow-ups)
- Starting a sentence with "Certainly"

CONTEXTS
CLIENT: Add one level of formality. Full sentences. No shorthand. Lead with context, then ask. Close with clear next step.
TEAM: Casual. Skip the pleasantries. Get to the point in line 1. Contractions, first names, occasional humor.
PARTNER/VENDOR: Warm but efficient. Mutual-value framing. Shorter than client emails.
OUTREACH (COLD): 3 paragraphs max. Lead with a specific hook (not generic flattery). One clear ask at the end. No "would love to pick your brain."

FORMAT NOTES
- Subject lines: I write them myself or you give me 3 options to pick from.
- Don't add a PS unless I ask for one.
- Don't bold random phrases in the body; I'll bold what matters myself.

Adjust every line until it reflects how you actually write. The more specific your prohibitions, the more useful this becomes.

---

Step 3: Load Your Style Profile

The instructions above capture what you know about how you write. But a lot of what makes your email voice yours is unconscious: the specific rhythm of how you hedge, the way you soften a direct ask, the exact words you reach for when you're being complimentary.

That's what a Style Profile captures. It analyzes your real writing, not your description of your writing, and extracts the patterns you'd never articulate.

Once you have a Style Profile, paste the Master Prompt section into your project instructions at the top. It goes before your baseline rules and acts as the foundation layer: your actual voice patterns as extracted from real data, with your explicit rules layered on top.

The combination is more reliable than either alone. The Style Profile handles voice fidelity. The context rules handle register.

---

Step 4: Prompt With Context Tags

The last piece is how you actually use the project. The goal is to give ChatGPT just enough context to produce a useful draft without re-explaining your voice setup every time.

A prompt template that works:

[CONTEXT]: [one-line situation description]
FROM: me
TO: [who + their role if relevant]
ASK: [what you need them to do or know]
TONE: [any specific adjustment from the context default]
LENGTH: [target, e.g., "under 100 words"]

Example:

CLIENT: Following up on a proposal I sent 5 days ago, no reply yet.
FROM: me
TO: Product director at a mid-size agency
ASK: Nudge them to respond without sounding desperate or pushy
TONE: Confident, not apologetic
LENGTH: Under 5 sentences

You don't need to include every field every time. For internal team emails where the context is obvious, "TEAM: Can you review the deck before Friday?" is enough; the project instructions fill in the rest.

---

Common Email Scenarios and How to Prompt Them

The follow-up that doesn't beg:

CLIENT: Second follow-up on [project/proposal], 7 days since last touchpoint.
ASK: Check if they're still interested / need more info. One clean door to walk back through.
LENGTH: 3 sentences

The "I need this by Friday" to your team:

TEAM: Need [deliverable] from [person] by EOD Friday for [reason].
ASK: Confirm they can hit it or flag early if not.
TONE: Direct, no pressure but clear on the deadline

The cold outreach to a potential collaborator:

OUTREACH: Reaching out to [person, role at company] about [specific angle, not generic]
ASK: 20-minute call to explore [specific thing]
HOOK: [the specific reason they should care; reference their work, not generic flattery]
LENGTH: 3 paragraphs max

The difficult internal message (bad news, course-correction):

TEAM: Delivery is slipping by [X days] because [reason].
ASK: Inform team + propose adjusted timeline
TONE: Candid, not defensive. Own the slip.
LENGTH: Short; this isn't the place for paragraphs

In each case, you're giving ChatGPT the situation and the parameters. Your project instructions handle the voice.

---

What This Setup Gets You

The point of this configuration isn't to automate email. It's to handle the mechanical part of drafting (the word choices, the structure, the register calibration) so you can focus on the judgment part: what to say and how much.

With a properly configured Email Project, you draft in seconds and edit in minutes instead of writing from scratch. The drafts sound like you because the project instructions are yours. The context rules mean you're not reconfiguring for every email type.

And for the emails that matter most, the ones where voice really counts, your Style Profile makes the difference between "AI helped me write this" and "this is how I write."

---

Building a ChatGPT Style Profile is the fastest way to get from "sounds like AI" to "sounds like me." You can also explore how ChatGPT Projects work for writers, see the complete ChatGPT Projects setup guide for deeper configuration options, or apply the same voice-loading approach to Claude's Projects; the principle transfers directly.