People open emails in 2026 because the inbox has already decided, on their behalf, that you’re worth a look.
There are of course multiple reasons why someone opens your emails – trusting who it’s from, having previously had a valuable experience with previous emails and intrigue created by the subject line.
And subject line testing still has a place – but it’s fighting for scraps of a decision that’s increasingly made before your subject line is even rendered.
What earns the open now is your accumulated relevance to that specific person, read by an AI layer sitting between you and them, and then confirmed or denied by a human who trusts that layer more than they trust you.
Why people get this wrong
Ask your CRM team what drives opens and you’ll usually get the same answer: better subject lines, better send times, better personalisation.
Every one of those levers assumes a human is scanning a list of subject lines and making a conscious, line-by-line choice. Nine times out of ten when I have a look at a brand’s send history, the tactics are still built JUST for that human.
The AI layer sitting in front of them never gets a mention in the strategy deck – and many marketers aren’t even aware of it fully yet.
The subject line was never really in charge
Even at its peak, the subject line was doing less work than most marketers gave it credit for.
It could earn a glance. It rarely earned engagement and trust on its own (glances are cheap, trust never is).
What’s changed is that a machine now sits in front of your subscriber and pre-judges the trust and relevance questions, working from a much longer memory than any subject line test ever had access to.
“Best practice” is describing a channel that’s already gone
Urgency words. Emoji in subject lines. Personalisation tokens that insert a first name and call it done.
Most of what still circulates as email best practice was reverse-engineered from an inbox where a human read every subject line in sequence, one after another, deciding as they went and these are still needed for the human layer.
But, the AI layer doesn’t work that way.
What’s really happening – your emails have two readers: humans and machines
Two systems now stand between your email and the person you sent it to, and neither of them cares what your subject line says.
1. Apple writes its own line to go alondside the subject line, and it isn’t your set preheader (AKA snippet/preview text)
On an iPhone with Apple Intelligence switched on, Apple generates its own one-line summary of the email and shows that instead of your preheader.
It’s written on the device, from your actual body copy (Apple’s model, not yours, decides what matters). It often ignores your carefully crafted preheader completely.
That line is historically has been built from whatever text sits first in your live copy – the part a human would actually read if they opened the message. But now, Apple decides this for you.
2. Gmail reads the whole email before your subscriber does
Gmail’s Gemini summary card goes further still. Once your email is opened, Gemini reads the full body and produces its own short summary, sometimes with suggested next steps, sitting right there in the reading pane above your actual creative.
It’s switched off by default for UK, EU, Swiss and Japanese recipients (worth checking before you rebuild your whole strategy around a feature half your list will never see). Where it is switched on, Gemini is effectively writing the second impression of your email (the one that lands after your subject line has already done, or failed to do, its job), built from your real content rather than your marketing copy.
The providers are scoring you before either of those layers even runs
Underneath both of those visible features sits something older and more decisive. Every major mailbox provider now weights inbox placement, and therefore your chance of ever being seen, on accumulated engagement history rather than any single send.
A subscriber who’s opened, clicked and read your last dozen emails gets shown your next one prominently. A subscriber who’s ignored your last six gets deprioritised, sometimes without ever touching a spam folder.
Isn’t that just deliverability by another name? Hasn’t sender reputation always mattered to some degree? What’s actually new here, then?
The difference is speed and granularity. This used to update over weeks. It now updates in near real time, per subscriber, weighted on relevance rather than volume alone.
A pattern I see repeatedly in CRM and email marketing reviews is a brand treating every name on the list as equally reachable, when the providers quietly stopped agreeing with that assumption months earlier.
The commercial impact
None of this is a technical curiosity. It’s revenue, and it compounds.
The brands that treat this as a subject-line problem will end up optimising the one lever that matters least – while the AI layer and the provider’s engagement scoring quietly decide the outcome regardless of what that subject line says.
The decision
Having worked directly on CRM programmes and email campaigns for the last 20 years and generating millions of pounds in revenue, when If I was rebuilding a sending programme for 2026, I wouldn’t start with subject lines at all (they would barely be a line at the bottom of my list in all honesty!).
I’d start by reading my own emails the way Apple’s on-device model reads them: strip the images, strip the layout, and look only at the plain text in the order it actually appears.
Then I’d have a proper look at what proportion of that live text is legal copy, navigation links and footer boilerplate. It’s often the majority of the words an AI system actually processes – and every one of those words is competing with your offer for the summary it produces.
It depends on your objectives, but for most commercial sends the fix is a re-ordering job or ensuring you have a single message focus
Put what’s already true about your offer where the machine reads it first, so it sees the same thing you want your human to see second, and don’t have anything else competing with that.
Email Marketing isn’t dead
It’s stopped being judged the way it used to be judged: by a human, reading a queue of subject lines, one at a time, in order.
It’s now judged by a system reading your actual relationship with that person, message by message, well before the human weighs in at all.
Chase the old inbox and you’re optimising a decision that isn’t being made there any more (however good your subject line testing gets).
Build for the one that’s actually happening – relevance, read early, read by a machine that’s seen your last twenty sends – and the open follows, along with the parts of the relationship that were always the real point.
Want to know whether your team would catch this shift early, or spend another quarter optimising subject lines for an inbox that’s already moved on?
Find out with the free CRM Priority Check, it takes about five minutes >