The short version: Gmail, Outlook and Apple Mail now generate an AI summary of your email before most people ever open it, which means your subject line and first paragraph are being read and judged by a machine, not a person. Open rates still look fine because the AI triggers an "open," but clicks and replies are quietly dying underneath. Fix it by front-loading the actual offer in your first sentence, cutting the clever curiosity-gap subject lines, and writing for a summariser first and a human second.
More on this here: How to Choose Between Email Marketing Platforms When You're Switching .
Something odd started happening to my open rates
For years my newsletter sat at a steady 35 to 40 percent open rate. Nothing spectacular, but healthy for a list built on real subscribers rather than bought data. Then, over about four months in late 2025, the open rate stayed exactly where it was, but clicks fell off a cliff. Three campaigns to a list of 12,400 people, open rate holding at 38 percent, click rate dropping from 4.1 percent to 1.6 percent. Same list. Same offer type. Same send time. Something had changed on the other side of the send button, not on mine.
The answer, once I went looking, was obvious and slightly unsettling. It wasn't people opening my emails less carefully. It was AI opening them for people, summarising the content, and letting the human decide whether to bother clicking through at all.
What's happening in the inbox right now
Gmail's Gemini integration now offers a one-tap "Summarise this email" option that sits right above the message. Outlook's Copilot does the same thing automatically for longer emails in a lot of business accounts. Apple Mail's on-device summaries, part of Apple Intelligence, show a condensed version of the email content directly under the sender's name in the inbox list, before the email is even opened.
That last one matters most for small business owners. It means your subscriber can read a two-line AI summary of your email in the inbox preview and decide it's not worth opening at all. Your open rate metric never even fires because plenty of these interactions happen without a genuine open. When they do open it, and the counting pixel does fire, they've often already absorbed the gist from the AI summary and skip straight past your carefully written intro to see if there's a discount code or a date.
This is not a small shift. It's the difference between writing for a person scanning their inbox and writing for a machine that extracts facts and discards tone.
The uncomfortable part nobody selling email courses wants to say
Most of what gets taught about email marketing, the curiosity-gap subject line, the story-first opening paragraph, the slow build to the offer, was built for a reading pattern that a meaningful chunk of your list no longer follows. An AI summary strips out your carefully built suspense and just says "20 percent off ends Friday" whether you wanted it to reveal that or not. All the emotional groundwork you laid in paragraph two gets skipped because the summariser already gave away the ending in paragraph one's substitute.
I used to write like Joe Pulizzi taught in his content marketing work, lead with a story, earn the CTA, treat the reader like someone worth a few sentences of build-up before you ask for anything. Some of those business lessons from Joe Pulizzi still hold for blog posts and long-form content. They hold much less well for a marketing email that an AI is going to compress into a single line before a human even sees your name.
That's uncomfortable to admit after years of teaching storytelling-first email. But pretending the inbox hasn't changed doesn't help anyone's click rate.
How I rewrote my emails once I understood this
I changed the structure of every marketing email I send. Here's the exact process, step by step, the same one I now use with consulting clients when we work through their AI marketing setup together.
- Step 1: Put the offer in the first sentence. Not the second paragraph. Not after a nice anecdote. First sentence. "Book a 30-minute AI audit this week and I'll send you the pricing breakdown I use with my own clients." If the AI summarises that sentence, the summary is still your actual message.
- Step 2: Write subject lines that describe rather than tease. "3 things I changed in my email strategy this month" outperforms "You won't believe what happened to my open rates." AI summarisers reward specificity because there's something concrete to extract. Vague, clever subject lines give the summariser nothing to work with, so it either ignores your line entirely or, worse, invents its own version of what you meant.
- Step 3: Use short numbered or bulleted lists inside the body. AI models handle structured content far better than dense paragraphs when generating a summary. A three-point list about your offer is more likely to be reproduced accurately than three sentences of connected prose.
- Step 4: Keep one CTA, stated plainly, near the top and repeated near the bottom. Multiple competing calls to action confuse a human reader and confuse a summariser even more. Pick one action. Say it twice.
- Step 5: Test open rate against click rate separately, every single send. If open rate holds steady but click rate drops over a run of three or four campaigns, that's your signal the AI-summary effect is active on your list. It's not a subject line problem you can A/B test your way out of with the old playbook.
What this means for the rest of your marketing, not just email
The same pattern is creeping into other channels. Instagram captions get summarised by Meta AI in some markets before a follower scrolls past. Google's AI Overviews summarise your blog post before anyone clicks through to your site. If you've read how Instagram built a marketing strategy that wins around visual storytelling and slow-build engagement, understand that even visual platforms are layering AI summary tools on top of content now, and the businesses adapting fastest are the ones putting the value proposition where it can't be missed, not where it's most artistically satisfying to place it.
Brands that have always led with clarity rather than cleverness are naturally better positioned for this. Look at how GoPro built their entire marketing strategy around user-generated proof rather than persuasive copywriting, the product does the explaining, the footage does the convincing. That kind of show-don't-tell content survives AI summarisation because there's less interpretive layer for a machine to flatten. The lesson translates directly to email: fewer adjectives, more facts an AI can't misrepresent.
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.
What still works, and where I haven't backed off
Interactive content is one place AI summarisers struggle, because there's nothing static to extract. A quiz, a pricing tool, or a calculator embedded in an email or landing page can't be reduced to two lines the way a paragraph can. That's part of why I've kept pushing clients toward tools like the ones covered in interactive calculators that convert visitors, because an email that says "use this calculator to see your ROI" forces a click that no summary can substitute for.
I also haven't stopped using Mailchimp, despite everything above. The platform itself hasn't caused this shift, the inbox providers have, and switching tools won't fix a structural problem. What I did change is how I brief campaigns inside it. If you want to see how a company builds durable email habits around a platform rather than chasing every algorithm shift, the way Mailchimp positions itself in its own marketing strategy is worth studying, because their whole pitch has always been about making small business communication simple rather than clever, which turns out to be the exact instinct AI-proofed email needs now.
The bit that surprised me
I expected reply rates to drop along with clicks. They didn't. Once I front-loaded the offer and cut the preamble, replies on my personal list went up slightly, from around 0.4 percent to 0.7 percent across four campaigns. My theory, untested but consistent with what I saw, is that people who did open the full email after skipping the AI summary were more decisive readers who'd already filtered themselves in. Fewer casual opens, more committed ones. Smaller top of funnel, better quality further down it.
That's the trade small businesses need to make peace with. You will likely see your top-line vanity numbers get messier before they make more sense. Open rate as a headline metric is becoming less trustworthy by the month, not because your emails got worse, but because the thing counting the "open" isn't always the person you're trying to reach.
A quick gut check for your own list
Pull your last five campaigns. Compare open rate trend against click rate trend separately. If open rate is flat or rising while clicks are falling, you're very likely seeing the AI-summary effect, not a content quality problem. Rewrite your next send using the five steps above before you touch your subject line testing again, and give it three sends before judging the result. One campaign won't show you the pattern. Three will.
Free resource: The Ad Account Naming and Structure Cheat Sheet.
Frequently asked questions
Does AI summarising my emails count as an open in my email marketing platform?
Often yes. When an inbox provider like Gmail or Apple Mail generates a preview summary, some tracking pixels still fire even though a human hasn't read the full email, which is part of why open rates can stay steady while genuine engagement, measured by clicks and replies, quietly falls.
Should I stop using curiosity-driven subject lines completely?
Not completely, but treat them as the exception rather than the default. Descriptive, specific subject lines give AI summarisers accurate material to extract and give human readers a clear reason to open, so lean on clarity for most sends and save the curiosity gap for the occasional campaign where you need it.
Which email metric matters most now that AI summaries are common?
Click rate and reply rate matter more than open rate right now. Open rate can be inflated or made meaningless by AI preview features, so track click-through and, if you send personal or relationship-driven emails, reply rate as your real signal of engagement.
Is this AI summary problem affecting all small businesses equally?
No. It hits list-based email marketing hardest, particularly newsletters and promotional sequences sent to Gmail, Outlook, or iCloud addresses, which cover a large share of most UK and US small business subscriber lists. Businesses relying more on SMS, direct social messaging, or phone outreach will feel it far less.
Related reading: The AI Sameness Problem: Why Your Small Business Marketing Sounds Like Everyone Else's and Why Your AI-Written Marketing Sounds Exactly Like Your Competitor's.
