ChatGPT Email Generator: Prompts, Examples, and Where It Falls Short

ChatGPT has become the default first tool most Virtual Assistants and Executive Assistants reach for when they need to draft an email fast. It is free, available, and produces coherent output in seconds. Used well, it removes the blank-page problem that was eating 3-8 minutes per reply. Used badly, it produces generic AI-tell output that clients get flagged for by their contacts.

The difference between those two outcomes is the prompt. This guide covers what a working email-generation prompt looks like, five templates for common email tasks, the failure modes to expect, and where specialized tools beat prompting ChatGPT each time.

This guide is part of our broader coverage of best tools for Virtual Assistants and Executive Assistants.

The 4-part prompt structure that works

Every good email-generation prompt has four components. Miss any one and the output goes generic:

1. Recipient context. Who is this email going to? What is their relationship to the sender? What do they already know?

2. Goal. What outcome does this email need to achieve? Not "write a follow-up." Something concrete like "get them to confirm the Wednesday meeting time."

3. Tone. How should this read? Formal, casual, warm, direct? Ideally with an example sentence in the target tone.

4. Prior context. What has already happened in this thread? What was said in the last exchange? What is the sender's typical voice with this recipient?

A full prompt template that works:

Write an email from [sender name and role] to [recipient name and relationship]. The goal is [specific outcome]. Tone should be [descriptor]. Example of the sender's typical voice: [one or two sentences from a past email]. Prior context: [what happened in the previous exchange, if any]. Keep it under [word count] words.

Any prompt shorter than this will get generic output.

Five prompt templates for common email tasks

Template 1: Follow-up on an unanswered thread

Write a follow-up email from [name] to [recipient]. It's the second nudge on a thread from [X days] ago about [topic]. The recipient hasn't replied. Tone: warm but persistent, not pushy. Sender's typical voice: [example sentence]. Prior context: original email asked about [what]. Keep under 80 words.

Template 2: Client meeting confirmation

Write a meeting confirmation email from [assistant name] to [external contact]. The meeting is [date, time, format]. Purpose: [why]. Attendees: [names]. Sender is the executive assistant to [executive name] and writes on their behalf. Tone: professional, warm, matches [executive]'s voice. Sample of [executive]'s voice: [example line]. Keep under 100 words.

Template 3: Rescheduling a meeting

Write a reschedule email from [name] to [recipient]. Original meeting was [date and time]. Need to move because [brief reason]. Propose three new times: [times]. Tone: apologetic but not excessive; brief. Sender's typical voice: [example line]. Keep under 90 words.

Template 4: Declining politely

Write a decline email from [name] to [recipient]. The ask was [what they wanted]. Reason for declining: [briefly]. Tone: warm, appreciative, decisive; do not open a negotiation. Sender's typical voice: [example line]. Suggest one alternative if appropriate: [what]. Keep under 100 words.

Template 5: Introducing two people

Write a warm introduction email from [name] connecting [Person A] and [Person B]. Reason for the intro: [why they should meet]. Both parties have opted in. Tone: warm, brief, sets clear context for both sides. Sender's typical voice: [example line]. Keep under 120 words.

Each of these templates produces usable output in 15-30 seconds. The output still needs a light edit before sending, but the blank-page cost is gone.

Where ChatGPT fails on email

Failure 1: Over-polished output. ChatGPT's default voice is more polished than most senders' actual voice. Casual founders get emails that sound like a professor. Direct executives get emails full of hedging language. The fix: always include a voice sample in the prompt and edit for tone before sending.

Failure 2: Generic openings. "I hope this email finds you well" is ChatGPT's default and is a well-known AI tell in 2026. Strip these openings from every draft. Better yet, tell ChatGPT to not use them in the prompt.

Failure 3: Repeating information the recipient already has. ChatGPT tends to summarize past context in the email itself. Recipients know their own context. The fix: prompt for brevity and specify that the recipient already knows the background.

Failure 4: Missing the specific action. ChatGPT often writes emails that describe a situation without asking clearly for the next step. Emails without a clear ask are emails that get ignored. Always name the specific ask in the prompt.

Failure 5: No memory across sessions. ChatGPT does not remember previous conversations by default. Every new email requires re-prompting with the sender's voice sample. For a VA or EA writing 30 emails a day across five clients, this is 30 prompt sessions.

When to reach for a specialized tool instead

For occasional email drafting where you're happy to prompt each time, ChatGPT is free and flexible. For high-volume, same-account work, specialized tools have three advantages:

1. Persistent voice learning. Tools like Replyf, the Gmail Chrome extension we build, learn each client's voice from their sent mail once. Every subsequent draft uses that voice without re-prompting.

2. Native integration with the inbox. Replyf runs inside Gmail. You open a new thread and the draft is already there in the client's voice. No copy-paste from ChatGPT, no context switching.

3. Multi-account handling. For VAs and EAs supporting multiple clients, Replyf tracks each client's voice separately. Switching to a different client inbox uses a different voice profile automatically. ChatGPT cannot do this without manual prompt-switching each time.

The right rule of thumb: use ChatGPT for one-off email tasks where prompting is fine. Use a specialized tool for the recurring same-account writing that a VA or EA does daily.

Related reading

For the underlying communication skills that make any AI email tool actually useful, see Client Communication Skills for VAs and EAs. For matching a client's voice specifically, see Brand Voice in Email. For the broader tools stack most VAs and EAs use, see our tools pillar.

Managing prompt libraries with Custom GPTs

ChatGPT Custom GPTs (available on ChatGPT Plus and above) let you save your email prompt templates permanently instead of pasting them fresh each time. For a VA or EA writing 20-30 emails per day, this alone saves 30-60 minutes of daily prompting overhead.

To set up an email-drafting Custom GPT:

1. Open ChatGPT, click your profile avatar, and select My GPTs → Create a GPT.
2. In the configuration tab, paste your standard instructions in the Instructions field: sender voice sample, tone preferences, ban list of words to avoid.
3. In the Conversation starters area, add your four or five most common email types: "Draft a meeting decline," "Write a warm intro," "Draft a follow-up nudge."
4. Save the GPT as private (do not publish publicly if it contains client-identifying details).

Every future email session opens with the executive's voice already loaded. You start each new draft with a fresh conversation but the GPT remembers the tone rules. This closes the biggest gap versus specialized email tools: persistent voice across sessions.

For working VAs and EAs who cannot use ChatGPT Plus due to client policy restrictions, the same pattern works with Claude Projects (Anthropic) or Gemini Gems (Google). Each has slightly different quirks but the core workflow is identical.