Gemini Prompts for Email Marketing: Copy-and-Adapt Templates

Published November 11, 2026 · 11 min read

Gemini prompts for email marketing work best when you treat Gemini less like a chat partner and more like a briefed contractor: give it a role, the source material, the task, the constraints and the exact output format. The templates below are built that way, and they assume you may be feeding Gemini a Google Doc brief or a Sheet of segments rather than typing everything into the prompt by hand. Copy them, swap the bracketed variables, and check the output before it goes anywhere near a send.

What Gemini changes about prompt writing

If you already have prompts that work in ChatGPT or Claude, most of them will port over. Three practical differences are worth adjusting for.

Long context. Gemini handles large pasted inputs well, so you can drop in a full campaign brief, a long product doc or a whole email thread instead of summarising it. That removes the compression step where detail usually gets lost.

Google Workspace as a source. You can point Gemini at a Doc or Sheet and have it read the content directly. This is the biggest workflow shift: your brief, your segment table and your brand rules can live where they already live, and the prompt just references them.

Literal instruction-following. Gemini tends to take structured instructions at face value. If you say "output a table with four columns," you usually get a table. That is an advantage, but it means vague constraints produce vague-but-confident output. Add one line telling Gemini where the context lives ("use the content of the attached Doc") and one line locking the output format.

Built-in AI inside an ESP is fine for a one-off draft. It is not a prompt library, and the library is what makes the output consistent across a team.

When these gemini prompts email marketing templates are worth using

Good fit: lifecycle and campaign copy at volume, ten subject line variants for an A/B test, adapting one approved email across three segments, turning a Docs brief into a first draft, and writing preview text that extends a subject line rather than echoing it.

Poor fit: legal, compliance and pricing copy; anything resting on a claim you cannot verify; and founder-voice emails where the whole value is that a specific person wrote them.

Treat every output as a first draft. You own the accuracy, the tone and every claim. If a prompt asks Gemini for a proof point and you did not supply one, the correct response is to leave a placeholder, not to fill the gap.

The prompt templates

Every template below shares one skeleton: role, context, task, constraints, output format. Keep that order. When output disappoints, the fix is almost always a missing constraint or an unstated format, not a longer prompt.

Campaign email from a brief

Role: You are a lifecycle email copywriter for [brand], a [category] brand.
Context: Use the attached Doc titled [brief name] as the source. Audience: [segment].
Goal: [single goal, e.g. drive first purchase]. Offer: [offer].
Task: Write one campaign email.
Constraints: 120-160 words in the body. Grade 7 reading level. No em dashes. Do not use the words elevate, unlock, delve, seamless. Do not invent statistics, testimonials or reviews. If you need a fact I have not given you, insert [NEEDS FACT: description] instead of guessing.
Output format: Subject line, preview text, then body with greeting, 2-3 short paragraphs, one CTA, sign-off.

Adapt when: the offer changes or the segment is new. Re-check the CTA wording every time — it drifts.

Welcome email, first in a sequence

Role: You are writing email 1 of a 4-email welcome sequence for [brand].
Context: The contact just signed up via [form/source]. They have not purchased.
Goal: Set expectations and deliver the promised [lead magnet / first value].
Task: Write email 1 only.
Constraints: Under 130 words. One CTA. No discount in this email. Plain, direct tone. No exclamation marks.
Output format: Subject line, preview text, body, CTA text, and a one-line note on what email 2 should do.

Adapt when: the signup source changes. A newsletter signup and a demo request need different first emails.

Cart or browse abandonment nudge

Role: You are writing a cart abandonment email for [brand].
Context: The contact added [product category] to their cart and did not check out. Time since: [hours].
Goal: Bring them back to complete the order without discounting.
Task: Write the email.
Constraints: Under 90 words. No guilt language, no "you forgot" framing. One CTA. Do not mention the price.
Output format: Subject line, preview text, body, CTA text.

Adapt when: you do want to include a discount. State the discount explicitly and add a constraint that it must appear once, not twice.

Re-engagement email for inactive contacts

Role: You are writing a win-back email for [brand].
Context: The contact has not opened or clicked in [timeframe].
Goal: Get a click, or a clean unsubscribe.
Task: Write the email.
Constraints: Under 110 words. Acknowledge the silence without apologising three times. Offer one clear reason to stay. Include a secondary text link to manage preferences.
Output format: Subject line, preview text, body, primary CTA, secondary link text.

Adapt when: the inactivity window changes. Also worth running a version with no offer at all to see whether the offer was ever the reason people stayed.

Product launch announcement

Role: You are writing a launch announcement for [brand].
Context: Use the attached Doc for product details. Audience: [segment].
Goal: Drive [action].
Task: Write the email.
Constraints: Lead with the problem, not the feature. 140-180 words. No superlatives (best, revolutionary, world-class). Only use facts present in the attached Doc.
Output format: Subject line, preview text, body, CTA text, plus a list of every factual claim you made and where in the Doc it came from.

Adapt when: launching to existing customers versus prospects. The proof point that lands is different.

Role: You are an editor tightening a newsletter section.
Context: Here is the draft: [paste].
Task: Rewrite it for clarity and pace.
Constraints: Cut at least 20% of the words. Keep every fact and every link. Do not add new claims. Keep the existing heading.
Output format: Rewritten section only, followed by a bullet list of what you cut and why.

Adapt when: the draft is already tight. Ask for a structural edit instead — reorder the paragraphs and flag anything that belongs in a different section.

Subject line template

Role: You are a subject line specialist.
Context: Email goal: [goal]. Audience: [segment]. Offer: [offer].
Task: Write 10 subject lines, one per angle: curiosity, benefit, urgency, question, number, plain, contrarian, social proof, seasonal, and one wildcard.
Constraints: Under 45 characters each. No emojis. Avoid these words: [your overused words]. No clickbait that the email does not deliver on.
Output format: A table with columns: angle, subject line, character count, one-line rationale.

Asking for character counts in the output lets you scan for truncation without counting by hand. The rationale column is what makes the batch reviewable — you can reject a whole angle instead of arguing line by line.

Preview text template

Role: You are writing preview text that extends a subject line.
Context: Subject line: [paste]. Email body: [paste or reference the Doc].
Task: Write 3 preview text options.
Constraints: 40-90 characters each. Do not reuse any noun phrase from the subject line. Do not repeat the CTA. No ellipses at the end.
Output format: Numbered list with character count after each.

The usual failure here is Gemini echoing the subject line back at you. The "do not reuse any noun phrase" line is the fix, and it works far better than "don't repeat the subject."

Body copy template

Role: You are a direct-response email copywriter.
Context: Audience: [segment]. Goal: [goal]. Offer: [offer]. One proof point: [fact you can verify]. CTA: [exact CTA text].
Task: Write the full email body.
Constraints: [word count] words. Short paragraphs, maximum 3 sentences each. No invented statistics, no fabricated testimonials, no customer stories. If you need a fact I have not supplied, insert [NEEDS FACT: description].
Output format: Greeting, body, CTA, sign-off. Then a separate list titled "Facts I needed but was not given."

That final list is the most useful part of the template. It tells you exactly where the draft is standing on nothing.

Sequence and automation copy template

Role: You are planning a lifecycle sequence.
Context: Trigger: [what starts it]. Audience: [segment]. End goal: [goal].
Task: Draft a 4-email sequence.
Constraints: Each email has one job. State the gap in days between sends. State the exit condition (what removes someone from the sequence). No email repeats another's angle.
Output format: A table with columns: email number, purpose, send timing, subject line, body outline, exit condition.

Use this before you build anything. If you would rather start from the workflow itself, Mailerflow's AI automation builder takes a one-sentence description of a workflow and drafts the trigger, waits, conditions and emails for review before activation — so you can paste the sequence summary in as that description, or use it to sanity-check the draft against your plan.

Feeding Docs and Sheets context into your prompts

This is where Gemini earns its place in a marketing workflow. Two patterns cover most cases.

Reference the source directly. Name the Doc or Sheet in the prompt and tell Gemini to use it as the only source of product facts. This keeps your brief as the single source of truth instead of a paraphrase of it.

Paste a structured extract when the source is large or messy. If a Sheet has forty columns and half of them are internal notes, paste only the columns that matter. Cleaner input, fewer silent misreadings.

A concrete example. Suppose your Sheet has columns for segment, offer, proof point and CTA. One prompt can produce a whole set:

Role: You are writing one email per segment.
Context: Use the attached Sheet. Columns: segment, offer, proof point, CTA.
Task: For each row, write one email.
Constraints: Same structure for every email. 100-140 words. Use only the proof point in that row. Do not carry a proof point across rows. If a cell is blank, insert [NEEDS INPUT: column name] rather than writing around it.
Output format: A table with columns: segment, subject line, preview text, body, CTA.

Three caveats. First, Gemini will not invent missing cells, so blank rows produce blank or generic copy — fill the Sheet before you run the prompt. Second, check that it did not silently merge two similar rows into one email. Third, do not paste raw customer data into a prompt unless your workspace settings and data policy allow it; use segment names and aggregate descriptions instead.

Keep the prompts that work in a shared Doc with a version note and the date each was last used. A prompt library with five entries that everyone actually uses beats a folder of forty nobody opens.

Personalization: what to put in the prompt vs. what to leave to the platform

Draw the line at what Gemini can know.

In the prompt: tone, segment framing, which benefit leads, formality level, sentence length, CTA wording, what to avoid. These are decisions you make once and reuse.

In the platform: merge fields, conditional blocks, and anything based on what a specific contact did. Gemini does not know that this person viewed a pricing page twice last week.

So the prompt should produce the reusable skeleton, and the platform fills the specifics. Write "lead with the benefit most relevant to [segment]" rather than "mention that they browsed the [product] page."

That division matters more as personalization gets deeper. Mailerflow supports AI personalization steps and website-behavior data through its JS SDK, plus dynamic segments built from fields, tags, engagement, events and website behavior — including AI segmentation from a plain-language description. The practical effect is that a prompt written for a segment can be targeted at a segment, rather than rewritten per contact.

Mistakes that make Gemini output unusable

No audience stated. The output goes generic. Fix: name the segment and one thing that distinguishes them.

No length limit. Gemini writes long by default. Fix: give a word range, not a maximum.

No banned-words list. You get em dashes, "delve," "elevate," "in today's fast-paced world." Fix: keep a running list of the words your team is tired of and paste it into every prompt.

Asking for one subject line. You get one mediocre option. Fix: ask for ten across named angles.

Letting it invent proof. A confident sentence with a made-up number is worse than no sentence. Fix: the [NEEDS FACT] instruction, plus a read-through before anything is scheduled.

Accepting the first draft. The second pass is where the quality is. Fix: ask Gemini to identify the two weakest lines and rewrite only those.

Pasting an entire brand guide. Three rules beat thirty pages. Fix: extract the rules that actually change the output.

Forgetting the output format. You get prose when you wanted a table. Fix: state the columns.

Reusing a prompt without updating the offer. The prompt still says last quarter's discount. Fix: date-stamp your prompt library entries.

Pre-send checklist: claims verified against a source, links correct, merge fields tested with a real contact, preview text not repeating the subject, plain-text version checked, and unsubscribe link present.

FAQ

Do these prompts work in ChatGPT or Claude too? Yes, with minor edits. Gemini tends to follow format instructions more literally and handles longer pasted context, so prompts written for it usually need trimming rather than expanding when you move them.

Can Gemini write emails in other languages? Yes, but state the language and locale explicitly, and have a native speaker review before sending. Machine translation of tone is where most of the damage happens.

How do I stop Gemini from sounding like AI? Give it a voice sample, ban specific words, set a sentence-length ceiling, and ask it to rewrite the two weakest lines. The banned-words list does more work than any instruction about "sounding human."

Should I keep prompts in a Doc? Yes, with a version note and the date last used. It is the difference between a team asset and a folder of one-off chats.

Can I use Gemini output directly in Mailerflow? Yes — paste it into the visual block editor or a reusable template. If you would rather start from a layout, the AI email builder generates a complete responsive design, subject lines and preview text from a description, and you can edit the copy from there.

Where this fits in the prompt management cluster

This article is the overview: the shared prompt structure, the templates, and how to feed Docs and Sheets in as context. The cluster goes deeper on individual jobs as the supporting articles publish — subject line prompts, sequence prompts, personalization prompts, prompt libraries and versioning, and how to evaluate AI output before it ships.

The structure carries across all of them: role, context, task, constraints, output format. If you learn one thing here, learn that order. It is what turns a prompt that sometimes works into one a team can rely on.

For more specific starting points, the B2B email prompt templates and the ecommerce campaign prompts apply the same skeleton to different jobs, and if your drafts keep coming back flat, the common AI email prompt mistakes covers the failure modes in more detail.

Describe your next email or automation in one sentence and let Mailerflow's AI build it — free plan, no credit card.