Your opens went up and your revenue went down, and the email team is calling it a win. It isn't. AI inbox summaries are reading your mail before your buyer does.
The largest inboxes now put an AI assistant between you and the reader. It summarizes threads, sorts what matters, and in some versions offers a digest of the day's mail so the person never opens most of it. I'm a digital futurist, and I've been saying for a while that the next gatekeeper wouldn't be a spam filter. It would be an editor. Now it's here, and it doesn't care how clever your subject line was.
What we know about AI inbox summaries so far, and what we don't
The biggest email provider began rolling out AI-powered summaries and an AI-sorted inbox in early 2026 [1], starting with English-language accounts in the US and expanding from there. Other inboxes have their own versions.
Here's the honest state of the evidence. No controlled study yet isolates what these summaries do to campaign results. The closest signal comes from one large analysis of more than two billion emails [2], which found open rates rising while click rates fell, with newsletter click rates sliding from about 4.0% to about 3.4%. The authors suggested AI summaries could be part of the cause and said plainly they couldn't confirm it.
That's the pattern you'd expect if machines are opening mail and summarizing it while fewer humans go deeper. It isn't proof. But it's enough to stop reporting opens as a success metric.
What the technician sees, and what the strategist sees
The technician sees opens up and calls the subject-line test a winner. Then the technician sees clicks down and blames the creative, or the offer, or the season. Both conclusions rest on a number the machine is now partly generating.
The strategist asks what the reader actually experienced. In more and more inboxes, the reader didn't see your subject line, your preheader or your hero image. They saw a summary: one or two sentences the assistant pulled or paraphrased from your email. If your email led with “Don't miss out on our biggest event yet,” the summary has nothing to work with. If it led with “The three-bedroom floor plans open to the public Saturday at 9, and the first ten buyers get the upgrade package,” the summary has a reason to exist.
The technician optimizes the envelope. The strategist writes the letter the machine will quote.
The Marine Corps trained me as a Combat Correspondent, and the discipline that stuck is the one that matters here: every story needs one clear thing a stranger can repeat. An AI summary is a stranger repeating your story to your buyer. Give it the one thing.
The subject lineused to be yourheadline. Now it'sthe sentence themachine keeps.
Everything else in the email is supporting evidence.
The mechanism: the Survivor Sentence
The Survivor Sentence is the single line in your email that states, in plain words, what the reader gets and why it matters now. It has four parts:
- A concrete noun. The thing itself: the report, the floor plan, the price, the date, the product.
- A reader outcome. What changes for them, said the way they'd say it.
- A time anchor. Why this week, today, before Friday. Real reasons only.
- No decoration. No exclamation points, no “exciting news,” no wind-up. Summarizers drop decoration first.
Put it in the first two sentences of the body, and echo it in the subject line. Summaries tend to lean on the opening and on clearly stated facts. Emails that bury the point under a story or a banner image get summarized as “a promotional email from” your brand, which is the polite version of the trash can.
How to rewrite your top flows around the Survivor Sentence is something we work through inside the iFix Lab.
The math (illustrative): when opens lie
These are modeled numbers for illustration, drawn from no real account.
Before. A brand sends 1,500,000 emails a week. Open rate 38%, click rate 2.4% of delivered, so about 36,000 clicks. Email drives $900,000 a month.
After summaries arrive. Say the reported open rate climbs to 46%, partly because summaries trigger opens nobody read. Clicks slide to 1.9% of delivered, so about 28,500. Revenue falls about 20%, to roughly $720,000 a month. The weekly report says opens are up eight points.
The wrong fix. The team runs more subject-line tests and picks winners on opens. Each “winner” is the line that best provoked the machine, not the reader. Clicks keep sliding.
The words fix. The team rewrites the top ten flows so each opens with a Survivor Sentence. Say clicks recover to 2.3%, about 34,500 a week. That's roughly $140,000 a month back, with no new tooling and no new list.
Check this against your own click data before you believe my numbers. The point is the direction: when opens stop meaning reading, the words at the top of the body carry the weight.
Opens went up eight points while revenue fell 20%. Each group is scaled on its own.
The vanity metric just got vainer
Cheap leads that don't convert, open rates that don't sell. Teams at every size have been fooled by a metric that looked good on the dashboard and meant nothing at the register. AI summaries make opens the newest member of that club.
Here's the uncomfortable part for agencies and email vendors. A lot of reporting decks lead with opens because opens are big and easy. If your vendor's monthly report still headlines open rate, ask them to lead with clicks, replies and revenue per thousand sent. If they push back, you've learned something about the report.
Writing the Survivor Sentence: before and after
Here's what the rewrite looks like in practice. These are examples, not anyone's real emails.
- Before: “You won't want to miss this! Our biggest event of the season is almost here, and we've got something special planned just for you.” After: “Our winter sale starts Thursday at 9am: every coat is 30% off for 48 hours, and members get early access tonight.”
- Before: “Big news from the team! We've been working hard on something we think you'll love.” After: “The new scheduling feature is live: your clients can now book and pay in one step, starting today.”
- Before: “Checking in to see how things are going with your account.” After: “Your trial ends Friday, and the reports you built will stay saved if you upgrade before then.”
Notice what changed. Every “after” names the thing, says what it means for the reader, and gives a real reason for the timing. None of them is clever. All of them survive being cut to one line by a machine that doesn't care about your brand voice.
This is uncomfortable for teams that pride themselves on creative subject lines. The good news is that the craft doesn't disappear. It moves from teasing the reader to telling them something worth knowing, fast. That's harder to write and far more valuable. Buyers have been asking for clarity through all the noise for years. Now a machine is enforcing it on their behalf.
Where the common advice breaks between $1M and $10M a month
At $1M a month, email advice is about subject lines: test curiosity against clarity, add the first name, keep it under 50 characters. Those tests still teach you something about your buyer.
Between $3M and $10M, the volume is big enough that the summary layer starts shaping results across whole segments. Curiosity-gap subject lines, the darling of every swipe file, are the first casualty. A summary can't summarize a mystery. It summarizes whatever concrete fact it can find, and if there's no fact, it says you sent a promotion.
A machine can'tsummarize a mystery.It can only skip it.
What changes once you're already past $10M a month
- Retire opens as a headline KPI. Report clicks, replies, conversions and revenue per thousand sent. Keep opens as a deliverability sniff test only.
- Write a Survivor Sentence standard. Every email template has a required first-sentence slot with the four parts. Make it a field in the brief.
- Audit your top flows quarterly. Welcome, abandoned cart, post-purchase, win-back. Read what an AI summary of each would say. If the summary is useless, the email is too.
- Give lifecycle and brand the same rule. At this scale, the newsletter and the automations reach the same person. Both need to survive the same editor.
- Test with replies. Ask a real question in the first lines of a few sends. Replies are the one signal no summary can fake.
The decision framework I'd use
Five questions for every email before it's approved:
- What's the one sentence? If the team can't write it, the email isn't ready.
- Is it in the first two lines? Not under the banner, not after the story.
- Would a stranger repeat it correctly? Read it aloud to someone who doesn't work there.
- Does the subject line echo it? Same promise, fewer words.
- What will we judge it on? Clicks, replies, revenue. Decide before it sends.
Questions senior buyers ask
Our open rates went up. Isn't that good?
Only if clicks and revenue rose with them. If opens rose while clicks fell, something other than a human may be doing some of the opening. Judge by what happens after.
Should we stop testing subject lines?
No. Stop picking winners on opens. Run the same tests and judge them on clicks and revenue.
Can we write for the AI directly?
You can write clearly, which is what the AI rewards. Plain facts, early. Gaming a summarizer is a short-lived trick. Clarity lasts.
Does this kill storytelling in email?
No. Lead with the Survivor Sentence, then tell the story. People who click want the story. The summary just needs the point.
What about design-heavy emails with little text?
They're the most at risk. If the message lives in an image, the summary has nothing to read. Put the key sentence in live text.
How fast is this spreading?
It started with a subset of accounts in early 2026 and keeps expanding. Plan as if most of your list will see summaries within a year.
Before you approve the next campaign
Take your best-performing email from last quarter. Write down what an AI summary of it would say in one sentence. If that sentence wouldn't make you click, rewrite the top of the email until it would.
That habit, reading the field before the field reads you, is what the iFix Lab is built on. The weekly Signal Briefing gives you what moved across the field, each item with a verdict (hype, watch or working) and a move, so you change the template before the summaries finish rolling out to your list. It's $97 a month, cancel any time, with a 14-day refund on month one. Join the iFix Lab.
Sources
- Google, Gmail is entering the Gemini era, 2026.
- Omeda, Email Engagement Report for Q2 2025, 2025.