EXPLORATIONS

Your Beautiful Image-based Emails Are About to Become AI Casualties

As AI-generated email previews (like Apple Intelligence) become more common, marketers need to adapt their strategies to ensure their emails still make an impact. Learn how to stay ahead in this AI-driven landscape.

By
Jon Uland
Published on
October 2, 2024

Imagine this: Your CEO is on the phone, and she wants to know why her company's latest email send arrived with the following disclaimer: "This message cannot be summarized."

When Apple Intelligence officially rolls out in a month, a lot of marketers at small and mid-sized DTC brands that still use image-based emails are going to be scrambling to get up to speed on email development best practices. 

But let’s back up for a minute. Last year, I wrote about why a HTML-first approach to email was superior in every way. Now I'd like to rehash the subject from a new perspective. 

AI technology has changed dramatically over the past 12 months, and so have email development norms. But the fact remains: HTML-first emails still reign supreme in this industry. Not only for their flexibility around design, device compatibility, and dynamic features (looking at you, AMP), but also now because they play nice with AI features.

And we know there will be many more of those on the horizon. 

Artificial Intelligence Only Increases the Need for HTML-first Emails

Not long ago, Apple announced its collaboration with Open AI to offer generative AI features to its customers. One of those features is an AI-produced preview of your emails (Google’s Gemini offers the same for certain business customers). The AI engine reviews the subject line, the preheader text, and the email body to produce a quick-look at the email content before the user even opens the message.

Email-savvy Apple Mail users were quick to dive into the practical application of this feature, and the results did not disappoint — depending on your point of view, that is.

The AI feature requires content to review and summarize, and the content it receives from the mail client is essentially unrendered HTML. It's not loading email images and interpreting what's on them; rather, the AI is looking at the raw code of the email that was sent and trying to generate an overview based on the content there.

Image-based emails in particular lack readable content, which prevents the AI from producing an overview and therefore throws an error message: "This message is too short to summarize." Here at Modular, our own testing also found significant discrepancies between the message content and the AI reviews (follow us on LinkedIn page to see the results as we start posting them).

I love this example because I think the AI preview text actually adds to an otherwise boring Subject Line/Pre-header Text combination. The main point is communicated by the summary, but the details and perhaps the why of this email are lost due to the presence of images where there should be live text.

This is just more evidence that marketing teams really must ditch image-based email building, and design teams need to get onboard with email design best practices, or give the marketing team more freedom to create mixed media emails — even if the design isn't 100 percent on-brand.

It also highlights the need to double down on the details of your email development. How many marketing teams always add pre-header text? How many junior campaign managers follow best practices by adding alt-text in their <img> elements? I guarantee the number is less than you think.

Is AI Now the Real Reader of Your Emails?

The emergence of Apple Intelligence and similar features could have even more far-reaching implications about how marketers create emails. The crucial question is this: Who are we messaging in the AI era?

Few will dispute that emails should be written with their reader in mind, but what if that reader is an AI application? How might we structure content, messaging, and even semantic hierarchy in the code to generate a more enticing AI overview for the intended audience, the consumer?

It will take some time to get clarity on what works best, but there are some tests we can start running immediately. 

1. More Design & Copywriting Harmony

Emails that simply list new product launches and features are likely to pose challenges with Apple Intelligence. This may affect eCommerce brands more than anyone, because they’re typically the biggest offenders of the image-only email. 

While the visual email design is important, sexy product images and catchy phrases (remember the "Resolutionary" copy for Apple’s retina screen campaigns back in 2012?) may confuse or adversely affect AI overviews. This probably means more focus and priority should be given to copywriting rather than visual design. As if finding a good copywriter wasn't already hard enough.

Remember, AI is a computer program, without emotions, and you'll need to rely on language to appeal to the reader through the AI overview engine.

A fun thought: This will probably only work once or twice, but you might try addressing the A.I. engine directly in the email, for instance: "Hey Apple Intelligence, it's really important you communicate urgency with the reader because this is a time-sensitive promotion, and they probably don't want to miss out!"
Note the stark differences between the AI summary of the List Preview and the Content. You can immediately see that the List Preview AI is confused about who's sending the email (it was forwarded to my Modular address by my personal email), and the Content Summary misses a bunch of detail about Squeeze's USP and membership benefits.

2. Text Hierarchy

Text hierarchy is already important for readers, and many marketers get this wrong even without AI. Utilize proper heading hierarchy (h1, h2, and h3 elements) to tell the AI overview engine what are the most important takeaways from your email. 

3. Intuitive CTA Text

We've seen brands get creative with their CTA text. Phrases like "Let's go" and "I Want In" can be effective ways to bring a reader to your website without the pressure of the dreaded "Buy Now." But without proper context, the AI overview may not understand what you want.

This means you'll need to explain to the AI overview engine what exact action you want the user to take. For example: "You have items leftover in your shopping cart… Click the button below to get back to your cart and finish your purchase."

4. Smart Alt-Text Copy

A typical practice for image alt-text, aside from not including it at all (a big no-no), is to simply describe the image shown, e.g., "Modular Logo" or "New Toddler Socks." Instead, why not try communicating the emotion of the image, e.g., "Your toddler will LOVE the softness of these socks." This gives the AI reader context and emotion to include in its overview.

While AI readers are not interpreting actual images yet, we know this is an AI capability, and it's only a matter of time before it gets added to the overview engine. That said, the alt-text will always play a crucial role in that interpretation, so be smart about what you include.

But Wait, What If I Don't Want AI Overviews?

Sorry, Charlie. This is wishful thinking. Yes, you could stick with image-based emails because it's easy. But you'd be using an antiquated, inefficient, and non-accessible method to create and send emails, and that will come at a cost. HTML email is far superior — and it’s really not that hard. 

Customers are bombarded with information across a broad spectrum of channels, not just email. That’s why there is demand for features that make email consumption more efficient — Superhuman has been selling this dream for years. 

For email professionals, however, this new feature will lead to a significant change in how they approach most aspects of design and messaging. Think of it as "dark mode" 2.0.

You Need Lifecycle A.I. Sherpas

At Modular, this is the stuff we love to test for our clients thanks to our tight-knit team of lifecycle experts, HTML email developers, designers, and solution architects.

Schedule a chat with us today — we'll even send you a YETI mug just for taking the call!

Jon Uland is Partner and Head of Operations at Modular. As a marketer and developer, Jon built his career creating sustainable marketing ecosystems for major brands and tech startups, including Peloton, Teachable, and Warner Music Group. Follow him on LinkedIn.