Ask any chatbot for a Black Friday email for a skincare brand and you get the same opener nearly every time, something about "glowing skin" and "the deal you have been waiting for." Thousands of brands send a version of that email every November. That is the core problem with AI email marketing: the tools are fast and useful, but their default output is the average of everything already written. We use AI every day. We just never let it have the last word.
Where AI saves real time
The best uses are the boring ones. Jobs where you need volume, a fresh angle or a quick sort through lots of text, and where a human still makes the final call.
Subject line ideas
Give the model the email, the offer and five subject lines that worked for this brand before. Ask for 30 options in that style. Most will be weak. Three or four will be worth testing, and one might be better than what you would have written alone. Then run the shortlist through the rules in our subject line guide and A/B test the top two.
First drafts
A first draft from AI gets you past the blank page. It works best when you feed it real material: the product page, the founder's notes, customer reviews, the brand voice document and two past emails the client approved. The more specific the input, the less generic the draft. Then you rewrite it. Usually that means cutting a third of it and replacing every vague claim with a real detail.
Segmentation ideas
Describe the store, the products and the data you have, and ask which segments might be worth building. AI is good at listing options you forgot, like buyers of one category who never tried the second, or customers who only buy during sales. It does not know your numbers, so treat the list as a brainstorm and check each idea against real data in Klaviyo. Our segmentation guide covers the segments we build in every account.
Review mining
This might be the most useful one. Export a few hundred product reviews, paste them in, and ask what customers mention most, which words they use for the problem the product solves, and which objections come up before purchase. You get the customer's own language, which makes the best email copy. A line like "finally a serum that doesn't pill under makeup" beats anything a copywriter invents. Check the quotes against the source before you use them, because models sometimes blend or reword them. More on collecting reviews in our review email guide.
Where AI email copy fails
Unedited AI copy has tells that readers notice even when they can't name them.
- Same vocabulary. Certain words show up in AI text far more than in normal writing. Once readers see them three emails in a row, your brand starts to feel like a template.
- Groups of three. "Soft, breathable and built to last." AI loves the triple. Real people rarely talk like that.
- Contrast setups. Lines like "it's more than a candle, it's a ritual" are everywhere now.
- Invented facts. Models will happily make up a statistic, an ingredient benefit or a customer story. In regulated categories like supplements or skincare, that can cause real trouble.
- No point of view. AI hedges and smooths everything out. A founder who has an opinion about their product sounds nothing like that.
So every AI draft gets a human pass. We read it out loud, cut everything that sounds like an ad, check every number against the brief and ask one question: would the founder actually say this?
Keeping the brand voice
AI will only sound like a brand if you show it what the brand sounds like. Before we use any model for a client, we write down how they talk.
| Give the model | Why |
|---|---|
| Voice notes with words the brand uses and words it never uses | Stops the default marketing tone before it starts |
| 3 to 5 approved emails | Examples teach style better than any description |
| Product facts and claims you are allowed to make | Reduces invented benefits |
| Real customer quotes | Gives the copy concrete language |
| A list of banned phrases | Catches the most common AI tells |
For brands in other languages, write the prompt and the output in that language from the start. Translating an English AI draft into German stacks two layers of "not quite right." A native speaker should still read every line.
Klaviyo's built-in AI
Klaviyo has added AI to several parts of the platform. Broadly, it can help with subject lines and email copy inside the editor, build segments from a description in plain language and show predictive data on customer profiles, such as expected next order date or likelihood to churn. Some of these need enough order history before they work.
We use them where they save clicks, especially the predictive data for timing and the segment builder for a quick start. For the copy itself we treat the built-in tools like any other model: a draft that needs editing. Features and names change often, so check Klaviyo's current docs for what your account includes.
A simple workflow that works
- Write the brief yourself: one idea, the offer, the audience, the real details.
- Use AI for angles, subject line options and a rough draft.
- Rewrite the draft by hand in the brand's voice.
- Check every claim and number against the brief.
- Read it out loud. Anything that sounds like an ad gets cut.
AI makes a good team faster. It won't fix a weak brief or a brand without a clear voice, and it can't tell you which email your customers will actually click. Your test results do that.
Questions we get a lot
Can AI write my marketing emails?
It can write a first draft fast. In our experience the draft still needs a person who knows the brand to cut the filler, add real details and check every claim before it goes out.
Which AI tool is best for email copy?
ChatGPT and Claude both work well for drafts and idea lists. The tool matters less than what you give it: brand voice notes, real product facts, past emails that performed and customer reviews.
Does Klaviyo have AI features?
Yes. Klaviyo has built AI into several parts of the platform, for example help with subject lines and copy, building segments from a plain description and predictive data on customers. Features change often, so check Klaviyo's current docs for what your plan includes.
Will AI-written emails hurt deliverability?
Not because AI wrote them. Deliverability depends on authentication, list quality and engagement. Bland emails that nobody opens or clicks hurt engagement over time, though, and that is a real risk with unedited AI copy.