ChatGPT Email Prompts: 32 Templates With Examples

Last updated: October 2026.

ChatGPT email prompts are reusable instructions that turn a blank page into a finished draft. Give ChatGPT a role, an audience, a goal and your [VARIABLES], and it writes quotes, booking confirmations, invoice chases and review requests in your voice. DigiFix uses these 32 prompts daily for Australian clients.

Every prompt below sits in a copyable block, uses [VARIABLES] so you can reuse it, and comes with a one-line note on why it works plus a filled Melbourne example. Prompts tested on 16 October 2026.

How do you write a ChatGPT email prompt that works?

Most people type a one line request, get a bland reply and blame the tool. A prompt that works names five things: the role ChatGPT should play, the reader, the single goal of the email, the tone, and the hard constraints such as word count, subject line and what must not be said. The table below is the framework DigiFix uses on every client sequence.

Here is the difference that framework makes. Both prompts ask for the same email.

Why it works: it removes every guess ChatGPT would otherwise make on your behalf, so the first draft is close to sendable.

Melbourne example: role set to office manager at Fitzroy North Plumbing, audience a Brunswick landlord with a burst pipe quote from 4 days ago, goal to book the repair this week, tone friendly and direct, under 130 words.

Which ChatGPT prompts help win more sales?

Sales emails are where ChatGPT pays for itself fastest, because the follow-up that never gets written is the one that loses the job. These seven prompts cover the moments where Australian service businesses most often go quiet.

Why it works: two concrete time options turn an open question into a simple choice, which is far easier to answer than a vague nudge.

Melbourne example: quote 4821 for a ducted heating service in Ringwood, sent 9 days ago, valid until 31 October, windows Tuesday 7.30am or Thursday 1pm.

Why it works: confirmations that state the preparation steps cut failed appointments, and the reschedule line keeps the booking out of a no-show.

Melbourne example: Hawthorn East unit, split system install, 14 November at 7.30am, two hours on site, clear access to the outdoor unit and a parking spot out the front.

Why it works: a same day recovery on a missed call keeps the enquiry warm, and admitting you do not know the reason stops ChatGPT inventing one.

Melbourne example: missed call from a Coburg homeowner at 4.52pm on a Friday, trades phone number and a booking link, sent before 6pm the same day.

Why it works: a real observation proves the email was written for them, and asking for 15 minutes instead of a sale lowers the cost of replying.

Melbourne example: a Dandenong North cafe with a Google listing showing no opening hours, offering a local SEO audit, 15 minute call on Thursday.

Why it works: it answers the objection with value and an option, so you keep the margin and still give the customer a way to say yes.

Melbourne example: a $4,200 bathroom renovation quote in Preston, inclusions listed as waterproofing, tiling and rubbish removal, with a phased option to do the tiling later.

Why it works: naming the referrer in the opening line is the strongest trust signal a cold email can carry, and it takes one line to include.

Melbourne example: referred by a Werribee accountant after a bookkeeping handover, offering Tuesday 10am or Wednesday 2pm.

Why it works: a yes or no question is the lowest effort reply in email, and the price or stock update gives a legitimate reason to make contact again.

Melbourne example: a Camberwell kitchen quote from 5 months ago, noting the supplier lead time has dropped from 8 weeks to 4, asking whether they are still planning the renovation this year.

What does this look like for a Melbourne trade business?

Take a two person electrical business in Glen Waverley turning over roughly $600,000 a year. The owner answers the phone between jobs, so around one enquiry in five goes to voicemail. Quotes go out by email and then nothing happens for two weeks. Invoices go unpaid for 45 to 60 days.

Here is the order DigiFix would work in, and how long each step takes.

Total build time is under three days. The prompts do the writing, but the sequencing is what changes the numbers, and that is the part most businesses skip. You can see the results DigiFix has delivered for other Australian businesses if you want a sense of the scale.

Which prompts handle complaints or reviews?

Service emails carry more risk than sales emails, because a badly worded complaint reply can turn a fixable problem into a public review. Four prompts cover the whole arc.

Why it works: naming the exact problem proves the email was read, and the action plus timeframe gives the customer something concrete to hold you to.

Melbourne example: a Clayton customer whose hot water unit was installed 3 days late, action is a $150 credit on the invoice, timeframe is the credit applied within 2 business days.

Why it works: the job specific detail makes it read like a human wrote it, and the warranty reminder reduces repeat call-outs you would absorb for free.

Melbourne example: a Northcote bathroom fan replacement, noting the existing ducting was reused, 5 year parts warranty, office number listed.

Why it works: suggesting what to mention produces reviews with the service keywords in them, and dropping any incentive keeps you inside Google’s review policy.

Melbourne example: sent 24 hours after a blocked drain cleared in Richmond, asking the customer to mention how quickly the team arrived.

Why it works: one clean apology with a new date reads as professional, while three apologies read as a business in trouble.

Melbourne example: a Werribee solar install moved from 12 to 19 November because the inverter shipment was held up, with the installer arriving first on site at 7am.

How do you chase overdue invoices politely?

Overdue invoice emails fail for one reason: they all sound the same at every stage. The fix is an escalation ladder where each email has a different job. Use the table to pick the stage, then the prompt.

Why it works: repeating the payment details removes the most common reason a first reminder gets ignored, which is that the customer cannot find the original invoice.

Melbourne example: invoice 1129 for $880 to a Bayswater client, due 2 October, bank transfer details repeated and an offer to resend the PDF.

Why it works: asking for a date rather than payment turns a stalemate into a commitment, and the hold makes the consequence real without a threat.

Melbourne example: invoice 1104 for $2,450, 19 days overdue, new work on hold, payment date requested by 30 October, 2 per cent monthly late fee per the terms.

Why it works: a stated process with a real date is defensible if the debt is disputed, and it removes the emotion that makes final notices backfire.

Melbourne example: invoice 1058 for $3,100, 34 days overdue, final date 7 November, referral to the business’s external collections process after that.

Why it works: a one word confirmation is the easiest possible reply, and a plan that recovers most of the invoice beats a debt that recovers none.

Melbourne example: a $1,800 invoice split into 3 payments of $600 starting 1 November, with a reply of agreed confirming the plan.

Why it works: stating what is not changing stops customers assuming the worst about the rest of the relationship, and advance notice protects the renewal.

Melbourne example: a managed IT retainer moving from $1,200 to $1,350 a month from 1 January because of monitoring licence costs, with the response time unchanged.

Why it works: naming the downstream customer booking moves the email up the supplier’s queue, and asking for a reply date stops the thread going quiet.

Melbourne example: PO 7742 for 12 LED high bays promised for 20 October, a Somerville warehouse install booked for 28 October, reply requested by 23 October.

Why it works: leading with the change rather than the background is what makes an internal email get read before the shift starts, not after.

Melbourne example: a new job photo requirement for the install team at a Campbellfield workshop, effective Monday, photos uploaded to the job app before leaving site.

Which prompts improve email marketing results?

Campaign emails are written once and sent to everyone, so a small wording change is worth more here than anywhere else. Seven prompts cover the full campaign cycle.

Why it works: asking for angles rather than options guarantees variety, and the ranking forces ChatGPT to justify choices you can then test.

Melbourne example: a Fitzroy cafe emailing a winter lunch special to 1,400 local subscribers, angles spanning curiosity through to a plain statement of the offer.

Why it works: preview text that repeats the subject line wastes the second most visible line in the inbox, and this prompt makes it add information instead.

Melbourne example: subject line about a EOFY stocktake sale for a Preston wholesaler, with preview lines that name the closing date.

Why it works: setting the frequency expectation in the first email is the cheapest way to reduce unsubscribes later, and the quick win earns the second open.

Melbourne example: a Hawthorn bookkeeping firm sending a monthly cash flow tip, a free BAS checklist, and inviting replies to the office inbox.

Why it works: asking for a table and a goal per email produces a plan you can build from, rather than four emails that all ask for the same thing.

Melbourne example: a Doncaster mortgage broker, 4 emails over 12 days moving from a rate update to a client story to a calculator to a booking link.

Why it works: the exclusion stops ChatGPT writing tax advice on your behalf, which is a genuine risk for any Australian business emailing in June.

Melbourne example: an equipment supplier in Dandenong offering a prepay discount for orders placed before 30 June, with delivery scheduled into July.

Why it works: a genuine local detail makes a seasonal email feel written for the suburb rather than pulled from a template library.

Melbourne example: a Brunswick cafe promoting a winter soup lunch during the coldest week of the year, running 1 July to 31 August.

Why it works: offering a reduced email frequency keeps the customer on the list, where a discount alone often just trains people to wait for the next one.

Melbourne example: a Nunawading pet supplies customer whose last order was in March, offering a repeat order reminder and a monthly instead of weekly email option.

How do you make ChatGPT sound like your business?

By default ChatGPT writes in a neutral, slightly American register. Three prompts fix that, and they are the ones worth setting up first because every other prompt inherits the result.

Why it works: three real emails give ChatGPT more voice information than any adjective you could type, and the banned list is reusable in every prompt afterwards.

Melbourne example: three quote emails from a Ringwood garden maintenance business, which produced a voice description of short sentences, first names only, and a habit of ending on the next step.

Why it works: locking the facts and shortening at the same time stops the rewrite drifting away from what the email actually needs to say.

Melbourne example: a 240 word draft from a Sunshine accounting firm shortened by 30 per cent, with the fee and the deadline left untouched.

Why it works: asking for the change list as well as the rewrite means you spot the patterns ChatGPT keeps repeating and can add them to your banned list.

Melbourne example: a draft from a Frankston retailer where organise, colour and 03/15/2027 were all corrected before sending to a customer.

What does Australian spam law require in your emails?

AI changes how fast you write an email. It does not change the law that governs sending it. The Spam Act 2003 applies to any commercial electronic message sent from Australia or received in Australia, and the ACMA enforces it.

Three obligations matter most for AI written emails. You must have consent before you send, the message must identify you as the sender with correct contact details, and it must include an unsubscribe option that is honoured within 5 working days without a fee and without requiring the person to log in (ACMA, Avoid sending spam). Consent can be express, such as a form or a ticked box, or inferred from a provable ongoing relationship where the marketing relates directly to it. Buying a list does not transfer consent to you.

Deliverability rules sit on top of the legal ones. Google’s Gmail sender guidelines, updated from 1 February 2024, require SPF or DKIM authentication, a spam rate below 0.3 per cent as reported in Postmaster Tools, and one-click unsubscribe on marketing messages, with additional requirements for senders of 5,000 or more messages a day (Google Workspace support).

Privacy sits alongside both. If your list contains personal information, the Australian Privacy Principles apply, and APP 7 sets the conditions for using or disclosing that information for direct marketing (Office of the Australian Information Commissioner).

Why it works: it turns a checklist into a step you actually run, and the instruction to defer on legal questions keeps ChatGPT inside its lane.

Melbourne example: a promotional email from a Springvale wholesaler missing an unsubscribe line, corrected to include the business name, ABN and an unsubscribe sentence.

Why it works: running the check on the prompt itself, not just the output, catches the details you paste out of habit while building the email.

Melbourne example: a draft quoting invoice 1104 for a named Camberwell customer, rewritten with [INVOICE NUMBER] and [CUSTOMER NAME] before it reached ChatGPT.

Do these prompts work in Claude, Gemini or Copilot?

Yes, with small adjustments. The five part framework is portable because it describes the task rather than the tool. What changes is how each assistant handles long instructions and how it stores context between sessions.

Why it works: the final line is the adjustment that matters most outside ChatGPT, and plain text output drops straight into any email client without reformatting.

Melbourne example: used in Copilot inside Outlook by a Geelong based sales rep to draft a quote follow-up, then sent with two edits.

One more thing worth knowing: how these assistants answer is still changing fast, and so is the way search engines surface content. If you want to understand where AI tools are heading before you build a process around them, read our guide to the limitations of ChatGPT. If your copywriting prompts are the gap rather than email, we cover that separately in the guide to the best ChatGPT copywriting prompts, so the two libraries do not overlap.

What else should you know about ChatGPT email prompts?

Are ChatGPT email prompts free to use?

Yes. Every prompt on this page works on the free tier of ChatGPT and on free accounts with Claude, Gemini and Copilot. You only need a paid plan if you want longer context, higher message limits or the ability to save custom instructions permanently. For a small business sending a few hundred emails a month, the free tier is usually enough to start. Upgrade when you find yourself hitting the daily limit during a busy period rather than before.

How many prompts do I need for a sales sequence?

Most Australian service businesses need four to six. A typical set is an initial reply, a quote follow-up at day three, a second follow-up at day eight, a closing email at day twenty one, and a booking confirmation once the job is won. That covers the full path from enquiry to scheduled work without repeating the same message twice. Adding more than six usually means the sequence has overlapping steps rather than genuinely different ones.

Will ChatGPT emails sound robotic?

They will if you skip the voice setup. Prompts 27 and 28 exist because a generic prompt produces generic writing, with the same rhythm and the same transitional phrases every time. Once you feed ChatGPT three real emails and ask for a voice description plus a banned word list, the output narrows to your register. The remaining tell is length. Most AI drafts run about twenty per cent long, so set a word limit and hold it.

Can I paste real customer details into ChatGPT?

Avoid it. Names, email addresses, phone numbers, addresses, invoice numbers and any financial or health detail should be replaced with [VARIABLES] before you send the prompt. A prompt is a message to a third party service, and free consumer accounts may use submitted content to improve their models. If your business handles personal information, the Australian Privacy Principles apply to how you manage it, and that includes what you paste into an external tool. Prompt 31 turns this into a quick check.

Does Google penalise AI written emails?

Google’s spam rules do not turn on how an email was written. They turn on whether the recipient consented, whether the message is relevant, and whether the sender is authenticated. What does get you filtered is a high spam complaint rate, which Google’s Gmail sender guidelines put at a threshold of 0.3 per cent in Postmaster Tools, along with missing SPF or DKIM records and a broken unsubscribe link. Write with AI, but send to a list that asked for your emails.

How long should a follow-up email be?

Under 130 words for a follow-up, and under 90 words for a second or third attempt. The pattern that works is short and specific: a reference to the original quote or booking, one new piece of information, and a single question or pair of time options. Long follow-ups read as a pitch and get deferred. If a follow-up needs more than 130 words to make sense, the problem is usually the original quote rather than the email chasing it.