The short version: Gmail's Gemini and Outlook's Copilot now generate AI summaries of your marketing emails right inside the inbox, which means people are deciding whether you're worth their time without ever opening what you sent. Open rate has been shaky since Apple's Mail Privacy Protection in 2021, and this finishes the job. If you're still reporting open rate as your headline email metric in 2026, you're reporting on something that barely exists anymore.
Next step on this topic: hope marketing strategy.
Next step on this topic: How to use Twitter Moments for Social Media Marketing (Even Though the.
What's changed in the inbox this year
Gmail rolled its Gemini summarisation feature out to more Workspace and personal accounts through 2025, and by early 2026 it's sitting in front of a huge chunk of your list without them ever turning it on. Outlook has its own version through Copilot. Both do the same basic thing: they read the email on the recipient's behalf and hand them two or three lines of "here's what this says" before the person decides whether to open it, skim it, or bin it.
That is a new problem for anyone sending marketing email, and almost nobody selling email marketing advice has caught up with it yet.
Think about what that means practically. Your subject line used to be the only gatekeeper. Now there's a second gatekeeper sitting behind it, an AI model deciding which three sentences of your 400 word email are worth surfacing. If your best line, your discount, your deadline, or your one persuasive sentence is buried in paragraph four, the AI summary might never mention it. The reader sees a flat, functional summary and moves on, having "read" your email without reading a word you wrote.
The uncomfortable number nobody puts in the case study
Here's the bit that's hard to say out loud if you sell email marketing services for a living: open rate has not been a trustworthy metric since September 2021, when Apple's Mail Privacy Protection started pre-fetching images for every Apple Mail user regardless of whether they opened the email. Overnight, plenty of small business lists saw open rates jump from a believable 22% to a suspicious 91%, because the tracking pixel fired automatically whether a human looked at the email or not.
Most agencies and most in house marketers quietly kept reporting open rate anyway, because it's the easiest number to put in a slide and clients like seeing big percentages. I've sat in enough review meetings to know that nobody wants to be the one who says "that 68% open rate means nothing." So they don't say it. The dashboards still have an open rate box, so it still gets filled in.
AI summarisation is the next layer of the same problem, except now it cuts both ways. Sometimes the pixel fires because Gemini pre-loaded the email to summarise it, so you get an "open" from someone who never saw your actual content. Other times the AI summary is good enough that the person decides not to open at all, so you lose the open entirely even though they absorbed your message. Either way, the number on your screen is disconnected from what a human did with your email. That's not a small technical footnote. That's your primary reporting metric being unreliable in both directions at once.
What happened with a client's list in real life
I work with a small homeware brand, a team of eight people selling candles and linen through Shopify, built up nicely over four years with a list of around 14,000 subscribers. Their open rate had sat between 68% and 74% for well over a year, which already should have raised an eyebrow given industry averages for retail email sit closer to 20-25% according to most of the benchmark reports agencies quote. Then in one week in mid 2025, it dropped to 34%.
The team's first reaction was panic. Had the list gone cold? Had they been marked as spam? We checked deliverability, checked sender reputation, checked nothing had changed in their send times or frequency. Nothing had. What had changed was that Gmail had pushed Gemini summaries to a larger slice of their subscriber base that month, and a chunk of opens that used to register as full opens were now being satisfied by a summary card instead.
Here's the part that mattered: click rate barely moved. Revenue per email barely moved. The number that fell off a cliff was open rate, and the numbers that measured actual behaviour, clicks and purchases, stayed almost flat. That told us the open rate had never been a good proxy for engagement in the first place, it had just been a comforting number that happened to correlate loosely with clicks until AI summarisation broke the correlation completely.
What to track instead of open rate
If open rate is unreliable, the fix isn't to ignore email metrics, it's to move your attention to the numbers AI summarisation can't quietly distort.
- Click to open ratio, which shows what percentage of the people who did open clicked something. This survives AI weirdness better because it's a ratio between two behaviours, not a raw count against a phantom pixel fire.
- Revenue per email sent, the number that pays your bills, tracked in your ecommerce platform or CRM rather than your email tool.
- Reply rate, on anything you send that invites a reply. AI summaries can't fake a reply. If someone writes back, a human read it.
- List growth versus list decay, tracked monthly, because a healthy list with falling opens is a very different problem to a dying list with falling opens.
- Landing page conversion from email traffic specifically, tagged with UTM parameters, so you can see what happens after the click rather than trusting the open count at all.
None of this is complicated to set up. Most email platforms already give you click to open ratio if you know where to look, and most ecommerce platforms will let you filter revenue by traffic source. The change isn't technical, it's just deciding to stop leading with open rate in your monthly report.
Writing emails that survive an AI summary
If a machine is going to compress your email into two lines before a human ever sees the whole thing, you need to write for that compression, not against it.
- Put the actual offer or point in the first sentence. Not the greeting, not the scene setting, the point. If the AI summary only grabs your first two lines, make sure those two lines contain something worth acting on.
- Use one clear number per email. A percentage off, a deadline, a specific figure. AI summarisers latch onto numbers because they're easy to extract and easy to compress. A vague "great deals inside" gives the summary nothing to work with.
- Name the action, don't imply it. "Reply with your size" or "book your slot before Friday" gives the summary an obvious verb to surface. A soft, brand-voice sentence that implies urgency without stating it gets flattened into nothing.
- Cut the preamble entirely. Every "hope you're well" and "as promised" line is space the AI has to wade through before reaching anything worth summarising. Delete the throat clearing.
- Test what the AI surfaces. Send yourself the email through a Gmail account with Gemini summaries turned on and read what it produces. If the summary doesn't make you want to open the full email, rewrite it.
This is the same discipline that good scannable web copy has always needed, the kind you'll see in how Grammarly builds its marketing around instantly obvious value rather than clever build up, or how Miro leads every piece of content with the specific problem it solves before it gets anywhere near a feature list. Email just caught up to a rule that good landing page copy learned years ago.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The part that's uncomfortable if you sell email marketing
If you run a small agency, or you're the one person doing marketing inside a small business, and email performance reports are part of what you charge for, this is worth sitting with. A big slice of the email marketing industry has been quietly reporting a number, open rate, that stopped being trustworthy in 2021 and is now actively misleading in a new way on top of the old way. Nobody wants to tell a client their headline metric has been close to meaningless for four years. I get why. It's an awkward conversation.
But the businesses that switch their reporting to click to open ratio, reply rate, and revenue per send now are the ones who'll still have credible numbers to show when a client eventually asks why their open rate looks strange, because Gemini summaries and Copilot summaries are only going to spread further through 2026, not shrink back. Getting ahead of that conversation is better than being asked about it in a review meeting you didn't prepare for.
Where this fits with the rest of your marketing
Email isn't the only channel where AI is inserting itself between you and the person you're trying to reach. Search results now have AI overviews summarising your web page before anyone clicks through, social platforms are testing AI-generated summaries of long posts, and customer service is increasingly a chatbot reading your knowledge base before a human ever speaks to you. The pattern is the same everywhere: write the important thing first, make the numbers and actions explicit, and stop assuming anyone reads top to bottom.
If you want a live example of a brand that's built its whole content approach around this kind of immediate clarity, look at how Pinterest structures every piece of content around one visual idea per post rather than a long build up, or how Red Bull leads with the single most shareable moment rather than context. Interactive tools do the same job in a different format, letting a visitor get to their own answer fast rather than reading your pitch about it, which is part of why I've written before about using interactive calculators to convert visitors who won't sit and read a page. The underlying shift is identical: assume compression, not attention, and write accordingly.
Small businesses that adjust their email writing and their reporting now will have a genuine edge over the ones still chasing a 70% open rate that means almost nothing. It's a small, unglamorous fix, but it's the kind of fix that moves revenue rather than just moving a vanity number on a dashboard.
Frequently asked questions
Is open rate completely useless now because of AI email summaries?
Not completely useless, but it's no longer reliable as a standalone number. Treat it as a rough directional signal at best and lean on click to open ratio, reply rate, and revenue per email as your real measures of whether an email worked.
Which email tools show AI-affected opens differently to normal opens?
Most major email platforms don't yet separate a human open from an AI summary pre-fetch, which is exactly the problem. Gmail and Outlook don't publish that data to senders either, so you can't currently filter it out. That's the main reason to shift weight onto click and revenue metrics instead of trying to clean the open number.
Should small businesses stop sending long emails because of AI summarisation?
Not necessarily shorter, but front-loaded. Put your main point, your number, and your action in the first two sentences, then use the rest of the email for the people who do open it fully. A long email with a strong opening survives AI summarisation better than a short vague one.
Does this affect cold outreach emails the same way as newsletters?
Yes, and arguably more, because cold outreach depends entirely on a stranger deciding to engage with no existing relationship to fall back on. If your cold email's first line doesn't survive being compressed into an AI summary, you've lost the reply before the person has even opened your message.
Related reading: The Best AI Tools for Small Business Owners (What I Pay For) and The Best AI Marketing Tools for Small Business in 2026 (And the Ones I'd Skip).
For the bigger picture, see my full guide to AI marketing.
