How to Write a Follow-Up Email Using ChatGPT in 60 Seconds

Why AI Follow-Ups Sound Like AI

The first draft of a follow-up email using ChatGPT sounds artificial because a vague request produces the most statistically ordinary email in existence. Ask for “a polite follow-up” and you get the average of every polite follow-up ever written — which is to say, the ones nobody reads.

Look at what the second version does that the first doesn’t. It names the thing, gives a reason the reader should care today, and asks for one decision with a deadline attached. The first version asks for nothing at all — which is precisely why it will be ignored a second time.

Most of the damage comes from a handful of stock phrases the model reaches for by default:

• “I hope this email finds you well”
• “Just checking in” and “touching base”
• “Per my last email” — reads as passive-aggressive to almost everyone
• “I wanted to reach out” as an opening clause
• “Looking forward to hearing from you” as a closer that requests nothing

Banning phrases explicitly is the single highest-value instruction you can add to an email prompt. It’s worth understanding how negative constraints change AI output generally, because the same trick fixes reports, summaries, and messages too.

The Recall-Reason-Request Method

The Recall-Reason-Request Method is a three-part structure for follow-up emails: remind the reader what the thread was about, give them a reason it matters now, and ask for exactly one thing. Every sentence that does none of those three jobs gets cut.

The method isn’t a replacement for the Role, Goal, Context, Format structure we use across the site — it slots inside it. The Logic System shapes your prompt; the Recall-Reason-Request approach we teach at PromptPeakAI decides what the finished email is allowed to contain.

The Request is where nearly everyone loses the reply. Three questions in one email is not three chances at an answer; it’s a task the reader postpones.

Write a Follow-Up Email Using ChatGPT

Writing a follow-up email using ChatGPT is three moves: gather your context, run one prompt, read the draft against the three jobs. On a familiar situation it really does land inside a minute. The first time, budget two — most of which you’ll spend copying context across.

Open the thread you’re chasing and copy the last message you sent, plus the date. Add two lines of anything the email doesn’t say — what was agreed verbally, what’s blocking, what changes if this slips. Raw and messy is fine; you are not writing here, you’re supplying facts.

Paste the block below into ChatGPT and fill the square brackets. Keep the rules section exactly as written — that’s the part doing the heavy lifting.

You are helping me write a short follow-up email.

My role: [your job title] at [your company].
Writing to: [name or role, and our relationship — client, manager, supplier].
What happened before: [paste your last email or two lines of context, plus the date].
The one thing I want: [the single next step — a yes or no, a date, a signature, a reply].

Write the email in this structure:
1. One line reminding them what this is about.
2. One line giving a reason it matters now.
3. One clear request — only one.

Rules:
- Under 80 words in total, plus a subject line of six words or fewer.
- Direct, warm, human English. Short sentences.
- Do NOT use "I hope this email finds you well", "just checking in", "touching base", "per my last email", or "I wanted to reach out".
- If a detail is missing, leave a bracketed gap like [date]. Do NOT invent facts, figures, or commitments.

Read the draft once and ask three questions: does it remind, does it justify, does it ask for one thing? Then fill any bracketed gaps yourself. Never send a draft with a bracket still in it — that’s the single most common way people get caught out.

If you send follow-ups daily, put the ban list and your job title into ChatGPT’s custom instructions instead of retyping them. The prompt then shrinks to context plus the one ask.

Three Follow-Ups You Send Most Weeks

Three situations account for most professional follow-ups: a proposal nobody answered, an approval stuck on someone’s desk, and the recap nobody wrote after a meeting. Each needs a different tone and a different ask, which is the only thing that changes between these prompts.

Write a follow-up email to a client who has not replied to a proposal.

Context: I sent [service or project] on [date]. We last spoke about [topic]. It has been [number] working days.
The one ask: tell me whether it is a yes, a no, or a later — by [day].

Rules: under 70 words. Warm, unhurried, not apologetic. Do not chase or over-explain.
Do NOT use "just checking in", "circling back", or "I hope this email finds you well".
Do NOT invent anything I have not told you.

Note what the tone instruction is doing: unhurried, not apologetic. Ask for “polite” and the model over-apologises, which reads as weakness. Ask for “firm” and it turns cold. Naming the feeling you want gets you far closer than naming a register.

Write a short email asking for sign-off that has stalled.

Context: I need approval on [document or budget] from [role]. It has been waiting since [date].
What happens if it slips: [the concrete consequence — a date moves, a supplier reprices, a team idles].
The one ask: sign-off by [day], or tell me what is blocking it.

Rules: under 70 words. Respectful and direct, never passive-aggressive. Lead with the consequence, not with my frustration.
Do NOT use "per my last email" or "as previously mentioned".

Turn these raw notes into a follow-up email that confirms what we agreed.

Notes: [paste your messy bullet points].
Attendees: [names or roles].

Format: one short opening line, then a list of decisions, then a list of actions with an owner and a date each.
The one ask: reply if anything below is wrong, by [day].

Rules: under 120 words. Factual and brisk, no praise or filler.
If an owner or date is missing from my notes, write [owner TBC] rather than guessing.

Two other common follow-ups deserve their own treatment rather than a variant here: chasing money and thanking an interviewer. Both carry consequences a generic prompt handles badly, so we cover chasing late payments without damaging the relationship and the post-interview thank-you email separately. If you need a full three-email sequence rather than a single message, that’s a sales follow-up cadence, which works differently again.

The last row surprises people. A thank-you email that asks for something turns a good impression into a small obligation — leave the request out and let the note do its own work.

Check the Draft Before You Send

Every follow-up email using ChatGPT gets the same ten-second check, and it is structural rather than stylistic. You are not reading for elegance; you are confirming the email contains three things and nothing else. A draft that fails the check is almost always missing the Reason.

When a draft comes back wrong, the fault is usually traceable to one missing line in the prompt rather than to the model having a bad day.

Rather than starting over, hand the draft back with a correction. Rewriting from scratch loses the parts that were already fine, and it’s slower.

That draft is close, but it reads like a template. Rewrite it:

- Cut it to [50] words.
- Delete every sentence that does not remind, justify, or ask.
- Keep my one request exactly as it is.
- Match this voice: [paste two sentences you actually wrote last week].

What You Can Safely Paste In

Writing a follow-up email using ChatGPT is the point where AI use collides with company data, because the context you paste is a real client, a real number, or a real internal disagreement. The good news is that a follow-up needs the shape of the situation, not the identifying details.

Swapping “Northgate Logistics” for “[client]” and “£84,000” for “[amount]” takes about five seconds and costs you nothing in output quality. Most IT and compliance leads will tell you the redaction habit matters more than which tool you picked.

On a personal account, chats may be used to improve OpenAI’s models unless you say otherwise. You can switch that off in Settings, then Data Controls, then “Improve the model for everyone” — OpenAI documents the exact path in its guide to disabling model training while keeping your history. On Business and Enterprise workspaces, workspace content is excluded from training by default.

The free plan is genuinely enough for this task, and paying more doesn’t change what’s safe to paste. Two honest caveats: message limits reset on their own schedule, and OpenAI’s Free tier FAQ notes that ads may appear on the free tier in some countries. If you’re weighing an upgrade for other reasons, we compare the free and paid plans in detail, and the wider question of whether ChatGPT is safe to use at work has its own guide.

Where ChatGPT Falls Short Here

ChatGPT drafts the email. It does not decide whether the email should exist, and that judgement is the part that actually protects your relationships.

There’s also a limit worth naming plainly: AI cannot make an unwelcome message welcome. If someone isn’t replying because the answer is no, a better-worded email won’t change that — a phone call might.

Start with the situation you send most often, save the prompt somewhere you can find it, and run it enough times that you stop reaching for the blank draft. The recap variant pairs naturally with post-meeting action summaries if meetings are where your week disappears.