Latest / The Newsletter Podcast with Fexingo: Substack, Beehiiv, and Email Newsletter Business / How Newsletters Use Gen AI for Content Personalization
Transcript
- Lucas: If you've been following the newsletter space for the past year, you've seen the flood of AI writing tools — but most of them are aimed at generating the whole newsletter, which frankly, readers can smell a mile away. Luna: Yeah, there is nothing worse than opening a newsletter that feels like it was written by a bot. But I've noticed some newsletters are using AI in a smarter way — personalization. Lucas: Exactly. And I want to talk about one specific newsletter that's doing it really well: The Browser. It's a daily newsletter that curates five articles worth reading. Been around for years. In January 2026, they started A/B testing ai generated summaries that are personalized based on the reader's interest tags. Luna: Interest tags — so they ask subscribers to check boxes like 'tech', 'science', 'history', 'business' when they sign up? Lucas: Right. And then the AI rewrites the summary for each article to emphasize the angle most relevant to that reader. So if you're tagged as 'tech' and one of the five articles is about both AI and literature, your summary might lead with the AI angle. A reader tagged 'history' gets a summary that foregrounds the historical context. Luna: That is subtle — they are not creating different articles, just different entry points. And the content itself is still curated by humans, so quality stays high. Lucas: That's the key. The curation is still done by a human editor — the AI only touches the framing. And their A/B test showed a 22 percent lift in open rates for the personalized version over the generic one. That's massive for a mature newsletter. Luna: Twenty-two percent is huge. But what about click-through rate? Because open rate is one thing, but if people open faster but don't click, it's just a vanity metric. Lucas: Fair question. They also saw a 9 percent increase in click-through rate — not as dramatic, but still meaningful. And more importantly, churn among subscribers who had been inactive for more than 60 days dropped by 15 percent. So the personalization is actually re-engaging people on the edge of unsubscribing. Luna: Alright, so that's an established newsletter with a dedicated editorial team. What about solo creators using Substack or Beehiiv — do they have access to this kind of thing? Lucas: Both platforms have rolled out native AI personalization features in the last few months. Substack launched something called 'Smart Summaries' in March — it lets writers tag different sections of their post and then the AI can customize the preview snippet per subscriber based on their reading history. Beehiiv has a similar feature called 'Adaptive Content' that's in beta since February. Luna: Adaptive Content — what does that do exactly? Lucas: It allows you to write conditional blocks within a single email. So if a subscriber has clicked on cryptocurrency articles before, they see a block about DeFi. If they haven't, they see a different block explaining basics. It's like mail merge on steroids. Luna: That sounds powerful, but also easy to screw up. What happens when the AI gets the personalization wrong — like showing a 'welcome back' block to someone who has been reading every day? Lucas: That's the big risk. If the data is stale or the tags are too broad, you end up with weird mismatches. I've seen a newsletter where I got a summary that said 'for our readers interested in marketing' — and I had never indicated that interest. It felt like a bug, not a feature. The trust erodes quickly. Luna: So the ethical line is: personalization should feel like a concierge, not a stalker. And it has to be transparent. Some newsletters put a note at the bottom: 'This summary was tailored for you based on your interests.' Lucas: I think transparency is non-negotiable. And you need to give subscribers control over their tags — let them update or delete them easily. The Browser does that. They have a link in every email that says 'update your interests' and it goes straight to a preference page. Luna: That is smart because it also gives you cleaner data. If someone takes the time to update their tags, they are telling you exactly what they want. Lucas: Exactly. And that brings us to the bigger point: personalization only works if you have the infrastructure to capture and maintain good data. A solo creator with a thousand subscribers can do this manually with a few segments. But once you hit ten thousand, you need automation. Luna: So what is the one metric you would watch to see if personalization is actually working — not just open rate, but something closer to real engagement? Lucas: I would track 'subscriber lifetime value per interest tag.' If you know that people tagged as 'finance' have a higher lifetime value than those tagged 'lifestyle', you can double down on personalizing content for that segment. It's a business decision, not just a content decision. Luna: That is a nice way to tie personalization directly to revenue. And if you are a paid newsletter, you can even test whether personalized previews convert free subscribers to paid at a higher rate. Lucas: Right. And that's where the real ROI is — not just keeping people from unsubscribing, but moving them up the value ladder. Personalized content can be the push that turns a casual reader into a paying subscriber. Luna: I want to try this with my own newsletter. But I'm worried about the time investment. How long does it take to set up something like Beehiiv's Adaptive Content for a weekly newsletter? Lucas: If your content is already structured into clear categories, it might take an afternoon to write the conditional blocks. The harder part is deciding on your segments — you don't want too many, because then the AI gets confused and so do you. Start with three: maybe 'beginners', 'intermediates', and 'experts' if you have a skill-based newsletter. Luna: That is a good starting framework. And you can always expand later. Lucas: Exactly. And if these marketing conversations have sparked something you have actually used in your own newsletter or business, it is worth mentioning that listener support is what keeps this show ad-free and independent. A couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo, if you have gotten something out of them. Luna: Yeah, it really does make a difference. Every contribution helps us keep digging into these specific angles that bigger shows skip. Lucas: Alright, back to personalization. I want to talk about something that I think will become a bigger issue in the next year: when personalization goes too far and starts deceiving the reader. What if the AI not only personalizes the summary but also rewrites the article itself to fit the reader's biases? Luna: That is a scary prospect. You could end up in a filter bubble created by the newsletter itself, not just by social media algorithms. Lucas: Right. There is a difference between curating content and manufacturing it. I think the ethical boundary is: the core content should be the same for everyone. Personalization should only affect the framing — the subject line, the summary, maybe a call to action. Not the substance of the reporting or analysis. Luna: I agree. And that is where human editorial judgment still matters. The AI is a tool, not a replacement for the writer's voice. Lucas: And that is probably the biggest takeaway from this whole episode: personalization is powerful, but it works best when it amplifies human curation, not when it replaces it. The newsletters that will win are the ones that use AI to serve their readers better, not to cut corners. Luna: Well said. So if someone wants to start personalizing their newsletter today, what is the first step? Lucas: Look at your current subscriber data. Do you have any tags, any segments, any information about what they want? If not, send a survey or a preference center link. Start collecting that data today. Then pick one small test — maybe just personalizing the subject line for two segments — and measure the results. You don't need to do everything at once. Luna: Start small, measure, iterate. That is a solid plan. Lucas: For next episode, I want to talk about something that came up in our listener survey: how newsletters are using AI to generate video summaries for social media. That is a whole other can of worms. Luna: Oh, that sounds like a good one. See you then.