Tom Slipkus | B2B/SaaS conversion copywriter

Apple Intelligence Is Rewriting Your Emails Before Anyone Opens Them

Apple Intelligence started replacing email preheader text with AI-generated summaries over a year ago, and most SaaS marketing teams either missed it entirely or made small adjustments (clearer subject lines, front-loading the value prop) and moved on.

And honestly, fair enough, because the preheader change by itself may not seem like a big deal. But it was the first sign of a much bigger trend that has real implications for how SaaS companies should think about their email strategy.

So what does the Apple update mean?

Basically, Apple's AI reads the first 100-200 characters of an email and generates its own inbox preview, replacing the original preheader text. With Apple Mail at roughly 58% of email client market share according to Litmus's tracking data, more than half the market is seeing an AI-generated summary instead of the original preview.

And that shift is already showing up in the data.

Bluecore's analysis across their retail client base found that email click-throughs from iOS devices dropped by up to 25% since the iOS 18 rollout. Listrak reported similar numbers: click-through rates down 18% on iOS and 12% on non-iOS. So while open rates may look fine, the actual engagement is most likely deteriorating.

In B2B specifically, this trend hasn't hit as hard. StoneShot's data shows desktop email is still heavily Outlook (60-70% of desktop opens), and mobile is only about 25% of total B2B email activity.

But "lower than 58%" is still a significant chunk of your list, and that number could grow with every update cycle.

Gmail is making similar moves, just not as visibly

Gmail's approach is different enough that it's easy to miss the connection. Gemini-powered summaries in Gmail are post-open (they show up after the recipient clicks into the email, not in the inbox preview), so they don't affect open rates the way Apple's change does. But Gmail is making a series of moves that, taken together, point in the same direction:

The features are different from what Apple is doing, but the direction is the same. Inbox providers are building systems that evaluate whether your email is worth the recipient's attention, and they're using AI to make that judgment.

The pattern underneath both changes

Deliverability has always been a technical problem. Authenticate your domain, keep your sender reputation clean, scrub your list, don't do anything weird with your IP, and the emails should usually get through.

That's still true, but it's becoming incomplete.

Both Apple and Gmail are now adding content quality evaluation on top of the traditional deliverability infrastructure. Validity's 2025 benchmark report says it pretty directly: AI is having unintended consequences on deliverability. And testing data from Unspam showed that despite a global deliverability health score of 86/100, only 60% of emails reached a visible mailbox location.

Technical delivery success now overstates real inbox reach by roughly 40%, which is a crazy number when you think about it.

What's filling that gap is algorithmic judgment about whether your content deserves the recipient's attention. Not just whether your authentication checks out, but whether the email itself looks like something the recipient would actually want to read. And that judgment is increasingly being made by AI systems that are very good at recognizing generic, templated, undifferentiated content (because that's exactly the kind of content they were trained on).

Which creates an inevitable collision. Marketing teams are investing in AI to generate email copy faster, while inbox providers are investing in AI to filter out exactly the kind of content that AI tends to produce. The two systems are working against each other, and the emails caught in the middle are the ones that sound like everything else in the inbox.

As Joe Cunningham wrote at MarTech, when anyone can generate an email campaign in seconds, the large language model alone isn't a differentiator. What differentiates is whether the offer, message, and copy were actually positioned to perform.

AI slop is a real word now, and inboxes are reacting to it

Merriam-Webster made "slop" their 2025 Word of the Year, which tells you something about how fast the cultural awareness shifted.

On the sending side, the trend is only starting to gain traction. The Litmus 2025 State of Email report found 49% of marketers now use generative AI for static copy creation, and AI image generation in email increased 340% year-over-year.

On the receiving side, it's not just the email providers, but the users as well who are noticing and adapting. An Exclaimer study found that 80% of consumers say overuse of AI in communication would make them consider switching brands, and 41% say AI-generated messages lack authenticity.

So you've got a situation where nearly half of all marketers are generating email copy with AI, inbox providers are building filters that bury generic or AI-sounding content, and the majority of recipients say they can tell and don't like it. That's a lot of pressure converging on the same point.

What actually survives the filter

If AI is getting better at identifying generic, templated content, the obvious question is what doesn't get caught in that filter.

The answer? Specificity. Writing that uses the exact language your customers use to describe their problems doesn't read like AI output, because AI doesn't have access to those conversations. An onboarding email that names the specific objection a new user has about data migration sounds different from one that says "get started in minutes," and it sounds different in ways that both humans and algorithms can pick up on.

The other thing that survives is structural intentionality. An email with a clear reason for existing, a specific ask based on where the reader is in their buying process, and a plan for what happens when they don't convert (which is most people) has a much better chance of standing out and being deemed worthy of attention. That kind of valuable and authentic message is much harder to fake, and inbox AI seems to be rewarding it indirectly by rewarding the engagement patterns it produces.

Deliverability used to be a technical issue, but now it's becoming a content quality problem, one that starts with whether your email was written with insights from real humans or generic phrases that sound just the same as everything else in the inbox. Inboxes today are getting good at gatekeeping what gets through, and that's a factor that will only become more important as it improves further.