ChatGPT Prompts for Email Newsletters Readers Love

A newsletter is the hardest email to prompt for, and I think that’s because it’s the only one with a relationship attached. A promo has an offer, a welcome email has a job, but edition #46 of a newsletter has nothing except the expectation you’ve built across the previous 45. Ask ChatGPT to “write my newsletter” and it fills that space with the only thing it has: generic enthusiasm about your industry, which is exactly the material your readers subscribed to you to avoid.

My fix is the same one behind every prompt in my library: stop prompting for the whole thing and start prompting for the structure, with your material in it. For newsletters, the structure I use is a rhythm of three beats - a story, a tip, a CTA - and the seven prompts below each own one piece of it. ChatGPT does the shaping. The week’s material, the opinions and the stories stay yours, because those are the parts a reader can tell were outsourced.

The numbers that set the rules

Three figures shape every prompt below, and I’d rather show them than assert taste.

You get about nine seconds. Litmus’s read-time data has the average time spent with an email falling year after year: 13.4 seconds in 2018, 10 seconds in 2021, 8.97 seconds by 2022. That’s the last time they published the figure, and nothing since suggests it recovered. So the prompts below are ruthless about front-loading: the story earns its length or gets cut, and the tip has to survive a skim.

The welcome email is your best-read edition ever. In GetResponse’s benchmark of 4.4 billion messages sent through their platform, newsletters averaged a 40.08% open rate while welcome emails hit 83.63% with a 16.6% click rate. Twice the attention. Prompt 7 exists because the first edition someone receives deserves its own treatment, not a recycled regular issue.

Don’t trust your open rate to judge any of this. Platform dashboards report opens inflated by Apple’s automated pre-fetching - the GetResponse 40% is a dashboard number. Brevo’s benchmark across 175,000+ senders puts the true open rate around 20.73% once that inflation is stripped. I judge newsletter changes on clicks and replies, and I’d hold you to the same.

The base prompt

Set this up once, in the same chat or project where your brand voice system prompt already lives:

You are the editor of [newsletter name], a [cadence] email for
[audience - who they are, what job they're trying to do].
The promise we made at signup: [what subscribers were told
they'd get].
Voice: [2-3 adjectives; one thing we never sound like].

Rules for every edition:
- one main story, one practical tip, one CTA. Never two CTAs.
- the reader gives us roughly 9 seconds unless the opening earns
  more. First 2 sentences carry the edition's best material.
- never invent stories, examples, numbers or customer details.
  Work only with material I paste. If a section needs material
  I haven't given you, ask for it instead of writing around it.
- short paragraphs, no throat-clearing, no "welcome to another
  edition of".

Wait for this edition's raw material.

Why it works: the one-story-one-tip-one-CTA rule is doing most of the labor. Newsletters die of accumulation - three updates, two links, a PS - and a model will happily accumulate forever. The anti-invention rule matters more here than in any other prompt I publish, because a fabricated anecdote in an ad is bad copy, but a fabricated anecdote in a newsletter is a lie told to people who trusted you with their inbox.

The seven prompts

1. The edition planner

Here is my raw material for this edition: [paste - notes, links,
a voice memo transcript, half-written thoughts, whatever exists].
Propose 2 edition plans. For each: the story (which piece of
material leads and why), the tip (the single most useful thing
in here), the CTA (one action, and what makes this edition the
right moment for it). Tell me what's missing from the material
if either plan is thin.

Why: this is the prompt I run first every time, and the last line is its best feature. A plan that’s thin on Tuesday is a newsletter that’s late on Thursday. Getting the “you haven’t actually got a story this week” verdict early means you go find one, instead of asking the model to inflate what’s there.

2. The story opener

The story for this edition: [paste what happened - the situation,
what you did or observed, how it ended, what you took from it].
Write 3 openings, max 4 sentences each: one that starts mid-scene,
one that starts with the takeaway, one that starts with the
mistake. Use only details I gave you. Keep my phrasing where
it's distinctive.

Why: the story is the beat readers remember, and it’s also where AI writing is most detectable, so the prompt is built to shape your account rather than generate one. “Keep my phrasing where it’s distinctive” is load-bearing - the model’s instinct is to smooth your voice into anyone’s, and the odd specific phrase is what makes a newsletter sound written rather than assembled.

3. The tip section

The tip for this edition: [the thing - a tactic, a setting, a
prompt, a resource]. Reader context: [what they'd use it for].
Write the tip section, max 120 words: what it is, the exact
steps or settings, and the one situation where it backfires.
No preamble about why tips matter.

Why: the backfire clause is the credibility move. A tip with a stated limitation reads like experience, and a tip without one reads like content. The word cap keeps the section skimmable inside the nine-second reality - a reader who only gets this far should still leave with the useful thing.

4. The curation block

Links worth sharing this edition: [paste URLs with one line from
me on why each caught my attention].
Write a links section: for each, one sentence on what it is and
one on why my specific audience should care. My reaction stays
in - edit it, don't replace it. Cut any link where you'd have to
guess the relevance.

Why: curation without commentary is a feed, and readers already have feeds. The rule that your reaction survives editing is what separates “here are five links” from “here are five links through my eyes” - the second one is the newsletter. Use this block instead of a second story, not on top of one.

5. The CTA

This edition's one CTA: [the action + where it leads].
Warmth check: this goes to [whole list / active segment], and the
last hard ask was [when].
Write 3 versions: direct (the ask, plainly), contextual (ties the
ask to this edition's story), and soft (for editions where the
list has been asked a lot lately). One short paragraph each, no
false urgency.

Why: the warmth check is the part most newsletter CTAs skip. A list is a bank account - story and tip deposit, CTA withdraws - and the model can’t know your balance unless you state it. Asking for three temperatures makes the judgment call visible: some weeks the soft version is the right one, and having it written stops you shipping the hard ask by default.

6. The edit pass

Full draft below. Edit it: cut it by 20%, flag any sentence that
sounds like a newsletter cliché, check the rhythm (story, tip,
one CTA - flag anything that snuck in beyond that), and confirm
the first 2 sentences carry the edition's best material. Show
the edit as a list of changes, not a rewrite.

Why: “list of changes, not a rewrite” keeps you the writer. A full rewrite pass quietly reintroduces the model’s voice everywhere, and you can’t see what changed. A change list makes every edit a decision you approve, which is also how you learn your own recurring bloat.

7. The welcome edition

A new subscriber just joined from [source] after being promised
[the signup promise]. Write their welcome email: deliver
[the immediate thing they signed up for], set the rhythm ("every
[cadence], you'll get..."), and include one best-of link from
[paste 2-3 past editions worth showing off]. One CTA: reply and
tell me [a real question you'll actually read answers to].

Why: the GetResponse numbers above say this email gets double a normal edition’s attention, so it should be written once, well, not generated per-send. The reply CTA isn’t a warmth trick - replies train inbox providers to treat you as wanted mail, and the answers are edition material for prompt 1. It’s the only email in your program that reliably gets both.

What I keep away from the model

The stories and the opinions. Not for sentiment reasons - for detection reasons. After enough editions, your readers have a model of you, and generic material breaks it the way a wrong voice on the phone does. ChatGPT in this workflow is the editor and the structure-keeper: it plans, tightens, varies and flags. The moment it starts supplying the experiences, the newsletter becomes the thing readers unsubscribe from politely, by archiving it unread, which no dashboard will clearly show you.

That’s also my answer to “can I automate the whole newsletter”. You can automate the assembly. The nine seconds of attention are being paid to you.

Subject lines are their own job

I deliberately left subject lines out of these prompts, because I’ve already published the subject line prompts with mobile truncation limits and A/B pair structure built in, and the newsletter campaign prompt there leads with the edition’s most interesting specific. Run prompt 1 here first, and the “most interesting thing” it surfaces is exactly what you feed that subject line prompt. If email is your whole job rather than one channel of it, my guide for email marketing managers covers where the newsletter sits among the flows, and the complete guide to AI marketing holds the wider workflow.

FAQ

How much time should this actually save? For me the honest win is consistency, not speed. The drafting hours drop, but the real change is that a thin week produces a short edition instead of a skipped one, because prompt 1 tells you early what you’re working with. Lists forgive short. They don’t forgive silence followed by a promo.

Should I tell readers AI helps write the newsletter? My rule: the structure and editing help doesn’t need a disclosure, the same way using a human editor never did. What would need disclosing is the thing this workflow bans anyway - stories and opinions that aren’t yours. Stay on the right side of that line and the question mostly dissolves.

What if my newsletter is pure curation with no story? Then prompt 4 is your main beat and the “through my eyes” rule is your whole product. I’d still open with one or two sentences of you - the week’s sharpest observation - because that’s the part of a curation newsletter that can’t be replaced by a feed, or by someone else’s ChatGPT.