Latest / The DTC Marketing Podcast with Fexingo: Direct-to-Consumer Brands, Shopify, and E-Commerce Growth / How DTC Brands Are Using AI-Powered Email Subject Lines
Transcript
- Lucas: I want to talk about something that sounds small but can move the needle a lot for DTC brands: email subject lines. Specifically, how brands are using AI to write them. Luna: Okay, I'm interested. Because subject lines feel almost trivial — but they're the gatekeeper for the whole email. Lucas: Exactly. If the subject line doesn't get opened, nothing else matters. And I came across a really specific case: Helix Sleep, the mattress company. They ran a test where they used an AI tool to generate subject line variants for their promotional emails. Luna: What did they find? Lucas: They saw a 14 percent lift in open rates compared to their human-crafted subject lines. That's not huge enough to be a miracle, but it's statistically significant and saved their copywriters hours of brainstorming. Luna: Fourteen percent is meaningful. What kind of AI tool were they using? Lucas: They used a gpt based model, trained on their past subject lines and performance data. So it wasn't generic — it learned their brand voice. They'd feed it the email content and ask for twenty subject line options, then the marketing team would pick the best three to A/B test. Luna: So the AI is generating ideas, but humans still curate. That feels like a smart middle ground. Lucas: It is. And the interesting part is what the AI tended to favour. The model leaned heavily toward personalization tokens — things like 'Hey, your mattress is calling' — and emotional triggers like 'Don't miss out on better sleep.' The human writers had been more focused on product features, like 'New cooling technology in our mattresses.' Luna: And the AI's approach performed better. That makes sense — personalization and emotion tend to drive opens. Lucas: By the way, if these marketing conversations have sparked something you've actually used in your own work — we'd love your support to keep the show ad-free. You can buy us a coffee at buy me a coffee dot com slash fexingo. Every bit helps us keep digging into stories like this. Luna: Yeah, it's a small way to say thanks if you're getting value. We really appreciate it. Lucas: So back to Helix. One thing that surprised me: the AI also generated some subject lines that were way too gimmicky — like 'Your bed is begging you to read this.' The human team rejected those because they felt off-brand. Luna: So the AI doesn't have taste. It can optimize for open rate in theory, but it doesn't know the brand's identity deeply. Lucas: Right. That's the key limitation. Another DTC brand I looked at — a skincare company called Curology — tried a similar approach but gave the AI too much autonomy. They ran a subject line that said 'You won't believe what this ingredient can do' — and it triggered spam filters because of 'won't believe' phrasing. Their open rate actually dropped. Luna: Ouch. So there's a dark side. The AI can produce language that feels spammy or clickbaity without understanding the consequences. Lucas: Exactly. That's why the human layer is critical. The best practice emerging is: use AI to generate a large pool of subject lines, then have a marketer filter for brand voice, spam triggers, and emotional resonance. Then test the top candidates live. Luna: And what about segmentation? Are brands using AI to tailor subject lines by segment? Lucas: Some are. A pet food DTC brand called The Farmer's Dog told me they use AI to generate different subject line variants for new subscribers versus repeat buyers. For new subscribers, the AI leans into education — 'What your dog's food really means.' For repeat buyers, it's more loyalty-driven — 'We saved your dog's favorite recipe.' They saw a 9 percent open rate lift on the segmented campaigns. Luna: That's smart. It's not just one-size-fits-all AI; it's AI plus segmentation. That seems like the real power. Lucas: Totally. And the tooling is getting easier. There are now Shopify apps specifically for AI subject line generation. But I want to be careful not to overhype it. The open rate lift is real, but it's usually in the range of 5 to 15 percent. It's not a silver bullet. Luna: What about the risk of the AI making the brand sound like everyone else? If all DTC brands use similar models, won't subject lines start to sound the same? Lucas: That's a real concern. I spoke with a marketer at a fashion DTC brand who said they stopped using AI for subject lines because every email started to feel generic. The AI kept suggesting phrases like 'You'll love this' or 'Trending now' — which are fine, but they don't differentiate. Luna: So the human touch is still the differentiator. The AI can help with volume and speed, but the voice has to come from people. Lucas: Exactly. And I think the smartest brands are using AI not to replace the writer, but to give the writer more raw material to work with. One technique I liked: a brand called Boll & Branch, the bedding company, uses AI to generate twenty subject lines, then has their copywriter combine elements from two or three of the best ones into a single line. So it's remixing, not just selecting. Luna: That's creative. So the AI becomes a brainstorming partner, not the decision-maker. Lucas: Right. And let's talk about measurement. How do you know if ai generated subject lines are actually working? The same way you'd measure any A/B test — but you need a large sample size because the lift is often modest. Helix tested on a 50/50 split of their list, which was about 100,000 subscribers per variant, to get statistical significance. Luna: That's a big list. Smaller brands might not have that volume. What can they do? Lucas: They can pool tests over time. Instead of testing one email, they can run a series of tests on the same segment and aggregate the results. Or they can test on a smaller segment first — like 10 percent of the list — and use the winner for the rest. That's a common approach. Luna: Makes sense. So let's say a DTC brand wants to start using AI for subject lines. What's the first step? Lucas: First, audit your past subject lines and their open rates. That data is the training material. Then pick an AI tool — there are options like Jasper, Copy.ai, or even ChatGPT with a well-crafted prompt. Start by generating 50 subject lines for your next campaign, have your team rate them, and then A/B test the top three against your usual approach. Luna: And do you recommend customizing the prompt with brand guidelines? Lucas: Absolutely. The more context you give the AI, the better. Include your brand voice description, words to avoid, and examples of your best-performing subject lines. One brand told me they include a list of banned words — like 'free' or 'guaranteed' — because those can trigger spam filters. The AI can respect those constraints if you tell it to. Luna: That seems like a good safeguard. So the AI is only as good as the input you give it. Lucas: Exactly. And I think that's the big takeaway for DTC brands: AI is a tool, not a strategy. It can save time and surface ideas you might not have thought of, but the brand's voice and the customer's trust have to be protected by humans. Luna: And the open rate lift is real, but it's not magic. It's about incremental gains. Lucas: Right. A 10 percent lift in open rates can compound over time if you're sending multiple emails per week. But it starts with a good subject line that respects the reader. Luna: I'd love to see a case study on long-term retention. Does AI help keep subscribers engaged, or does it just boost short-term opens? Lucas: That's the next question. I think we need more data on that. But for now, the evidence shows it's a net positive — as long as you keep the human in the loop.