Newsletter prompts for ChatGPT and Claude

These 14 newsletter prompts help you plan, write and improve a newsletter with ChatGPT, Claude or Gemini: choosing what each issue covers, drafting from your own notes and links, writing subject lines and preview text that match the content, segmenting your list and reviewing what worked. Each prompt starts from your material, because a newsletter only earns its place in an inbox when it says something the reader can't get elsewhere.

How to use these prompts

Replace the blanks in double braces, or keep the default after the colon: {{cadence:weekly}} means "weekly" unless you change it.

Keep three things ready: two or three past issues (for voice and format), your notes and links for the next issue, and basic numbers from your email platform (opens, clicks, unsubscribes by issue). The prompts below ask for them.

These prompts are for recurring newsletters. For one-off campaign emails, welcome sequences and win-backs, see the marketing prompts; for everyday work email, the email writing prompts.

If you send in the United States, the FTC's CAN-SPAM Act compliance guide applies to commercial email, and it says plainly: "The subject line must accurately reflect the content of the message." Several prompts below build that in.

Which prompts help you plan a newsletter?

1. Define the newsletter's promise

A clear promise makes every later decision (topics, length, subject lines) easier.

I'm starting or relaunching a newsletter.

About me: {{about:a backend engineer who writes about running Postgres in production}}
Who I want to read it: {{readers:engineers at startups who own the database but aren't DBAs}}
How often: {{cadence:weekly}}
What I can realistically produce each issue: {{capacity:one main piece and a few links}}

Write 5 one-sentence promises for the newsletter ("Every week, you get
X so you can Y"). For each, list the recurring sections it implies and
one risk (too broad, hard to sustain, already well covered). Recommend
one and explain why.

2. Issue ideas from what you collected

Turns a messy pile of saved links and notes into a backlog.

Here's everything I saved over the last {{period:month}}: links,
notes, reader replies and questions.

{{saved_items}}

My newsletter's promise: {{promise}}

Group these into 6 to 10 issue ideas. For each: a working title, the
main point, which saved items support it, and what I'd need to add
(an example, data, my own opinion). Drop items that don't fit the
promise and list them separately.

3. A repeatable issue format

A fixed structure helps readers skim and makes writing faster.

Here are my last {{count:3}} issues:

{{past_issues}}

Propose a repeatable format: section names, order, target length for
each, and what goes in each. Base it on what already works in these
issues, not a generic template. Keep the total reading time under
{{minutes:5}} minutes. Then show an empty skeleton I can copy.

Which prompts write an issue?

4. Draft an issue from notes

The main writing prompt. Your notes and opinions in, a draft in your voice out.

Write this week's issue of my newsletter.

Format:
{{format}}

My notes and opinions for this issue:
{{notes}}

Links to include, with why each matters to me:
{{links}}

Match the voice of this past issue:
{{past_issue}}

Use only facts from my notes and links. Where I gave an opinion, keep
it and keep it mine; don't soften it. If a section has no material,
write [NOTHING THIS WEEK] instead of filling it.

5. Curated links with real commentary

Link roundups are only useful when you say why each one is worth a click.

Here are {{count:6}} links for the roundup section, with my rough
notes on each:

{{links_and_notes}}

For each link write 2 or 3 sentences: what it is, the single most
useful idea in it, and my take from the notes. Don't describe content
I haven't summarized for you. Order them from most to least useful for
{{readers:engineers who own the database}}.

6. Cut a draft to a reading time

Long issues get skimmed. This trims without losing the point.

Here's my draft issue, about {{current_words:1,800}} words:

{{draft}}

Cut it to about {{target_words:900}} words. Keep the main argument,
every concrete example and my voice. Remove repetition, throat-clearing
intros and anything that doesn't serve the main point. After the edit,
list what you cut so I can put back anything important.

7. A clearly labeled sponsor section

When an issue carries an ad, the reader should know it's an ad.

Write a sponsor section for my newsletter.

Sponsor: {{sponsor}}
What they offer: {{offer}}
Facts I can state: {{facts}}
My honest experience with it, if any: {{experience:none}}

Label it clearly as sponsored at the top. Under {{words:80}} words. Use
only the facts given, no claims I haven't verified, and don't imply I
use the product if my experience is "none".

Which prompts write subject lines and preview text?

8. Subject lines that match the issue

Clickbait gets opens once and unsubscribes after. Under CAN-SPAM it can also be a deceptive subject line.

Here's this week's issue:

{{issue}}

Write 8 subject lines. Each must accurately reflect what's in the
issue; no promises the issue doesn't keep, no fake "Re:" or "Fwd:",
no false urgency. Mix styles: the main takeaway, a specific number from
the issue, a question the issue answers, and the topic plainly stated.
Keep each under {{chars:50}} characters and show the count.

9. Preview text that adds to the subject

Many inboxes show a line of preview text beside the subject. Use it to add information, not repeat it.

Subject line: {{subject}}
Issue summary: {{summary}}

Write 5 options for preview text, each under {{chars:90}} characters,
that add something the subject line doesn't say (a second topic, a
detail, who it's for). Don't repeat the subject's words.

10. A subject line test plan

Tests are only useful if you know what you're testing and when the result counts.

I want to A/B test subject lines on my newsletter.

List size: {{list_size:4,000}}
Typical open rate: {{open_rate:38%}}
My platform's test options: {{platform_options:split a sample, send winner after 4 hours}}

Suggest one hypothesis to test this month (for example, numbers versus
no numbers), 2 subject lines that differ only on that variable, and how
to read the result. Tell me honestly if my list is too small for the
difference to mean much, and what to track instead (clicks, replies).

Which prompts help you segment your list?

11. A segment plan from your subscriber data

Start with the fields and behavior you already have.

Here are the subscriber fields and behavior data my email platform has:

{{fields:signup source, topics chosen at signup, last open date, clicks by topic, customer yes/no}}

My newsletter: {{promise}}

Suggest 2 to 4 segments I could actually write for. For each: the rule
that defines it using my fields, roughly what share of the list it
might be (ask me if you can't tell), what they'd want more or less of,
and one change to the issue for them. Don't suggest segments that need
data I don't have.

12. One issue, two versions

Small changes for different readers, without writing two newsletters.

Here's this week's issue:

{{issue}}

Segment A: {{segment_a:new subscribers in their first month}}
Segment B: {{segment_b:long-time readers who click most issues}}

For each segment, suggest a different intro paragraph, a different
subject line, and which section to move to the top. Keep the body of
the issue the same. Explain each change in one line.

13. Re-engage inactive subscribers

Inactive readers hurt deliverability. Ask them, and make leaving easy.

Write a short re-engagement email for subscribers who haven't opened
in {{months:6}} months.

My newsletter: {{promise}}
Best issues from that period: {{best_issues}}

Under {{words:120}} words. Remind them what they signed up for, link
3 of the best issues, and give two clear choices: stay subscribed or
unsubscribe with one click. Tell them that if they don't respond by
{{date}}, they'll be removed. No guilt and no tricks.

Which prompts help you review results?

14. Monthly performance review

Find out what readers actually value, not what you assumed.

Here are stats for my last {{count:8}} issues: subject line, topic,
send day and time, open rate, click rate, unsubscribes, replies, and
top clicked links.

{{stats}}

Tell me: which topics and formats get the most clicks and replies (not
just opens), any issue that caused an unusual number of unsubscribes
and a likely reason, and 3 changes to try next month. Be clear about
what is a pattern and what could just be noise with numbers this small.

What does CAN-SPAM require of a newsletter?

CAN-SPAM covers commercial email. The FTC's guide says "It covers all commercial messages" and "The law makes no exception for business-to-business email." When a message mixes commercial and other content, the guide says "the primary purpose of the message is the deciding factor." If your newsletter promotes products or carries ads, treat these as your checklist:

The guide says each separate email in violation "is subject to penalties of up to $53,088" (as of October 2026). It also notes that "subscribers and members don't lose their ability to opt out of marketing emails from you."

Inbox providers add their own rules. Google's email sender guidelines say that if you send more than 5,000 messages a day to Gmail accounts, "Marketing messages and subscribed messages must support one-click unsubscribe." Google also recommends keeping spam rates in Postmaster Tools below 0.10%.

This is a summary, not legal advice. Other countries have their own rules, so check the ones where your subscribers live.

How do you keep a newsletter written with AI worth reading?

Bring the substance yourself. Readers subscribe for your judgment: what you noticed, what you tried, what you think is overrated. Put that in your notes and let the model handle structure and polish.

Keep your voice samples current. Paste a recent issue into every drafting prompt. Anthropic's prompting best practices say "Examples are one of the most reliable ways to steer Claude's output format, tone, and structure."

Check every link and fact. Models can misdescribe an article they haven't seen. Only include links you've read, and make the model work from your summaries.

Measure clicks and replies, not just opens. An open only tells you the email was opened; a click or a reply tells you the issue was useful, which is why the review prompt weighs them more.

Keep your newsletter prompts in one place

A newsletter is a weekly habit, so the same five or six prompts get used every issue. Kipt keeps them as Markdown in a private GitHub repo you own, with their blanks ready to fill; tag them by step (plan, draft, subject, review) and press ⌘K to find the one you need. To turn your issue format and voice rules into a reusable skill for Claude or Codex, paste prompt 4 into the free SKILL.md generator. If you also share each issue on LinkedIn, the LinkedIn prompts help turn it into a post.