Insight
AI Email Summaries Are Rewriting Your Customers' Inboxes
13 May 2025 Customer Experience Management
You’ve crafted the perfect email. Sharp subject, clear layout, punchy CTA. And then your customers’ email provider skims it, summarises it, and tucks it into a promotions tab. Welcome to the AI-powered inbox. Whether or not AI email summary features are actually helpful, they’re here. A combination of technical and content tweaks – front-loading content, using specific language, relying less on layout, and getting familiar with structured data – keeps your emails alive.
New AI-powered features are quietly changing the way your customers see your emails (or don’t!).
Apple Mail, Gmail, Yahoo, and Outlook have introduced AI email summary features. An AI model reads an incoming email, decides what it’s about, and summarises it above the main body in 30–150 words.
AI also generates previews, reorders content, ignores subject lines, and filters messages into tabs. Which means your carefully planned customer journey might never land the way you intended.
AI is very much changing how we at Tap CXM build email templates. We’ve had a wave of clients asking the same questions: What’s happening? What should we do about it? Are open rates dead?
While it’s not a drop-everything emergency, I wanted to share some AI email summary optimisation tips that are working for our clients.
With the help of my email consultant colleagues, I’ve pulled together a quick, clear guide to help you adapt your email strategy for the new inbox without scrapping what already works.
- What changes are live now (and what’s coming next)
- How summarisation varies across platforms
- Why your design-first approach may need a rethink
- The smart, simple shifts that still give you control
Download the Guide
Practical takeaways you can share easily with your team.
AI Isn’t Decimating Email Engagement – Yet
We’re all familiar with the impact of AI summaries on Google search results. The situation in email marketing is a bit more nuanced.
A recent analysis of 2B+ emails found the average click-through rate dropped from 1.8% to 1.74% (unique click-through fell from 4.35% to 3.93%). However, opening rates climbed about 2.5%.
In other words, more opens and fewer clicks. Pretty much what you’d expect with AI summarising emails.
Engagement won’t fall off a cliff as long as you make emails AI-friendly
A bit of context
Apple is the world’s largest email provider with 64.66% of opens (including automatic MPP opens). As of Q1 2026, the company has shipped an estimated 450M+ devices with “Apple Intelligence” AI email summary capability.
How AI Actually Reads Your Email Marketing
AI doesn’t read emails the way it reads in-browser content.
For one thing, it doesn’t render HTML. It strips the email down to text and looks for signals that it can use to understand, categorise, and summarise your content.
Three things then happen in the recipient’s inbox:
- Summarisation: AI reads the body and writes its own preview. If you open with a brand paragraph before the offer, AI may surface the offer out of context or miss it entirely.
- Extraction: Specifics like codes, prices, and expiry dates are highlighted, e.g. Gmail creates badges. If those details are in an image rather than text, AI won’t summarise them and your customer won’t see them.
- Prioritisation: Engagement signals decide whose email gets a prominent slot. If yours is slipping, it gets quietly buried, with no warning in your analytics.
The final result can look different for different users, based on their email history and AI preferences.
Essentially, you no longer control what lands.
This is the same story we’ve told about AI becoming the customer’s new front end. Email is simply the next channel where you optimise for the machine as well as the human.
AI summaries won’t have full penetration for some time thanks to eligibility requirements, geographical exclusions, operating system requirements, and users opting out. Still, it’s going to happen at some point. Developing good templates and habits now puts you ahead of AI models and the brands that aren’t paying attention.
Hasan Jamil, Senior Technical Consultant
How to Manage AI Email Summaries
First, Find People Who Understand AI
A good AI strategy is built around human tasks and roles, but AI can’t and shouldn’t replace human skills.
You need people who understand how AI “thinks”, what it “likes”, and where it’s heading.
This field is moving fast and the rules are still forming. We’re encouraging clients to think about embedded AI skills like context engineering and confidence with AI agents, rather than hiring tech specialists for the sake of it.
Having people who keep an eye on trends and think critically helps you adapt faster than a first-class technical whiz.
Front-Load Important Information
Bring your value proposition upfront. Don’t make users scroll for it.
Buzzstream’s analysis of AI email summary results found that Google, Apple, and Microsoft included more content from early in the email.
The front-loading was especially strong in Google’s and Apple’s case, with 87% and 82% respectively coming from early in the email.
Google and Apple also showed the shortest summaries, indicating that length and prioritisation could be related.
This is the “AI fold” in email marketing. Front-load emails with the valuable stuff above the fold and save the brand voice and storytelling for further down.
Hasan Jamil, Senior Technical Consultant
Think in Entities
Using specific, entity-based language makes it easier for AI to analyse and parse your emails.
An entity is a uniquely identifiable thing or concept:
- Products
- Prices
- Dates
- Events
- Brands
- Offers
- People
- Places
Models keep a database of entities. As they read your email, they’re looking to extract entities and map how they relate.
“Get rewards for re-ordering” gives AI next to nothing. “10% cash back if you re-order before June 30 using the code SUMMERLOVE” gives it specifics and steps it can extract.
Just remember: AI-friendly, not AI-first. Email marketing should still convey a good amount of brand personality to connect with customers. Create for humans first. Optimise for AI second.
Hasan Jamil, Senior Technical Consultant
The Technical Layer: Structured Data, Not Pretty Structure
Email AIs read text and code. They don’t understand visual layout elements without clear instruction.
Tables, columns, and design don’t work. What does work is marking up the email’s underlying code.
Semantic Markup and Image Tags
Descriptive tags help AI models understand what they’re looking at.
There are two relatively simple tags that make a big difference:
- Semantic markup: HTML tags like <p>, <h1>, <h2>, and <ul> (for lists) define what the text is, not just what it looks like.
- Image tags: Clear, descriptive alt tags give AI important context clues for content it can’t see.
Checking code is easy in most big-name email automation platforms. Just be careful not to mess with universal elements like headers, footers, button styles, or dynamic content blocks unless you’re absolutely sure it’s the right thing to do.
When in doubt, call on your in-house email developers or reach out to us for a technical platform audit.
Schema.org Structured Data
Most top-ranking web pages use some form of schema. Almost no email marketer does.
Adding structured data as JSON-LD in the email’s code gives machines clear instructions on how to interpret the information.
A few schema types that work well in email:
- Organization keeps your brand and logo rendering correctly in previews (note US spelling is the standard in schema markup).
- DiscountOffer triggers deal badges in Gmail and stops AI from mangling your offer details.
- PromotionCard and its availabilityEnds property give a timed offer a second chance to appear at the top of Gmail’s promotions: once at send, once near expiry.
This will keep evolving as the models do. The key takeaway is that schema markup is a valuable lever that hardly anyone is leaning on in email marketing.
If you’re familiar with page graphs and structured data in SEO, it’s similar to that. This level of schema markup is standard on most high-ranking websites but underused in email. Brands that get good at it now have an advantage.
Hasan Jamil, Senior Technical Consultant
Why You Need to Optimise for AI Email Summaries Now
AI email summaries are relatively new. And the solutions are still largely manual.
But there are other reasons not to wait.
Almost every email provider has rolled out algorithmic “Most Relevant” email sorting. These algorithms analyse incoming emails, recipient engagement patterns, and sender reputation information to decide what is “Most Relevant” to the user.
Optimising for AI gives your emails a better shot at the top spot.
What You Can Do Today
A few practical moves you can make to get ahead of AI email summaries include:
- Send test emails to see how Apple Mail, Gmail, and Outlook handle your content.
- Use AI tools to summarise different variations.
- Start with schema markup on high-priority offers and expand to other key details.
- Keep templates lean (Gmail clips messages over ~102KB).
As far as I know, there aren’t any dedicated tools to test AI email summaries. It’s manual for now. Although it might change as AI summary capabilities roll out wider.
So, Is AI Coming for Email?
AI isn’t killing email. It’s changing how customers receive it. Which means it’s changing how marketers need to think about building emails.
Depending on how your email campaigns are developed today, optimising for AI email summary features might involve a few minor tweaks or a major overhaul.
Either way, don’t lose sight of the customer on the receiving end. Always create emails for customers first and optimise for AI second.
And if you need help getting up to speed, Tap CXM’s email marketing specialists are ready to jump in. We embed experts into your team to solve CXM problems and give you the retained knowledge to continue moving with the times.
