A client came to me in February with a deliverability puzzle. Their open rates had dropped 18% compared to the same period the previous year, but their click-through rates were nearly unchanged. The list was clean. Deliverability was healthy. The content hadn't gotten worse. What changed?
After digging into the data, the answer was uncomfortable: their subject lines had been optimized for an inbox that no longer exists.
The old model was simple. Subject line gets a human's attention. Human opens. Human reads. Human clicks. The subject line's job was to win the open. Write something intriguing, create a sense of urgency, maybe add a number or an emoji, and the metric would move.
Gemini broke that model. Not dramatically or all at once, but steadily across late 2025 and into 2026. A growing share of Gmail users now see a one-line or two-line AI summary of your email before they decide whether to open. The subject line is no longer the gateway. It's one signal among several. And the techniques that used to drive opens in the curiosity-gap era often work against you now.
What Gemini Actually Does to Your Subject Line's Role
When Gemini generates a summary card beneath a subject line, subscribers are reading a condensed version of your email's content before they open it. For many emails, especially newsletters and promotional messages, that summary answers the implicit question: "Is there anything in here worth my time?"
If the summary satisfies that question, some subscribers will read the summary and move on without opening. If it raises a question the summary can't answer, they're more likely to open. If the subject line and summary together signal that the email isn't relevant, they'll delete or archive faster than before.
This means the subject line now has a dual job. It signals value to the human scanning the inbox, and it signals content type and intent to the AI that will summarize the email. These two jobs are not always in conflict, but the techniques optimized for one can hurt the other.
The curiosity-gap subject line ("You won't believe what we found...") creates tension by withholding information. That worked in a pure open-rate game. But Gemini's summary reveals the content anyway, which deflates the curiosity-gap trick. Worse, vague subject lines give the AI less signal to work with, which can produce less accurate or less compelling summaries.
Why Specificity Performs Better Now
The subject lines that held up best through the Gemini rollout share a common trait: they are specific enough that both a human and an AI can accurately classify what's inside.
A clothing retailer I worked with ran a natural experiment across three subject line styles over Q4 2025:
- Curiosity-gap variant: "We need to tell you something about your spring wardrobe"
- Benefit-vague variant: "Big savings inside this week"
- Specific variant: "30% off all linen pieces: ends Sunday April 13"
The specific variant outperformed on open rate by 12% over the curiosity-gap version and by 8% over the benefit-vague version. CTR was 22% higher than the curiosity-gap version. The sample was 43,000 subscribers split equally across the three versions.
The reason specificity wins is structural. Gemini's summary is built from the email body, not the subject line. But the subject line primes the reader's expectation. When the subject line says "ends Sunday April 13," the summary card that mentions the sale reinforces that message. The human reading the inbox gets two consistent signals, and consistency builds trust that the open is worth it.
When the subject line is vague and the summary is specific (because the body had the real offer), there's a slight dissonance. The subscriber thinks: "Why didn't you just tell me that in the subject line?" That dissonance doesn't kill performance, but it's friction.
What Gemini Appears to Reward in Subject Lines
Based on the pattern of performance data I've seen across clients and what's known about how large language models classify and rank content, a few factors appear to matter.
Clarity about content type. Does the subject line immediately communicate what kind of email this is: promotional, informational, transactional, event invite, survey? Gemini's classification of email type is now sharper than it was with the older tabbed-inbox algorithm, and subject lines that match their email's actual type seem to get better placement and more accurate summaries.
Specificity signals. Numbers, dates, names, and product references help the AI anchor the summary to concrete claims. "Save 30% on running gear this weekend" gives Gemini three specific hooks: the discount amount, the product category, and the time frame. An AI-generated summary can work with that cleanly.
Absence of manipulation signals. Excessive capitalization, multiple exclamation points, words like "FREE" in all caps, and false urgency phrasing appear to trigger filtering behavior more aggressively now than in 2024. This isn't new, but the threshold has tightened. Subject lines that used to squeak through are now getting flagged more often.
What Gemini Ignores
This is where I'll disagree with a lot of standard email marketing advice.
Emoji are nearly irrelevant as a performance lever in Gmail. They do not appear to meaningfully influence Gemini's summary generation or inbox placement. They may marginally help in non-Gmail clients where they add visual distinction in the inbox. But the obsession with finding the right emoji combination for subject lines is solving for a variable that no longer moves the needle much where roughly half of most B2C lists live.
I've run the A/B tests. A SaaS company ran 12 subject line tests over six months with emoji variants. The average lift from adding emoji was 0.8% on open rate, and four of the twelve tests showed no lift at all. The time spent optimizing emoji choice is almost certainly better spent on the content claim in the subject line itself.
Curiosity-gap constructions also appear to get less benefit of the doubt from Gemini's classification system. "What no one tells you about email marketing" is a format that trained human readers to open. AI systems classify it as low-specificity, which can result in a less informative summary, which reduces the total signal value of the inbox row.
The Case for Treating Subject Lines as Headlines
Good subject lines in 2026 look less like teaser copy and more like good journalistic headlines. A newspaper headline doesn't withhold. It tells you what happened and trusts that the story is interesting enough to hold you.
"Fed Raises Rates 0.25%, Third Increase This Year" tells you exactly what happened. You either care about that or you don't. The headline doesn't try to manufacture interest through withholding. It surfaces the specific claim and trusts the reader to self-select.
Email subject lines work better with the same approach. "Your Q1 Email Report: Opens Down 14%, Here's Why" tells a subscriber exactly what's inside. They either want that or they don't. But the people who want it will open at a higher rate than if the subject line had been "Something important about your email performance this quarter."
This specificity-first approach requires knowing what the actual value of your email is before you write the subject line. Which is a good discipline anyway. If you can't write a specific subject line, it's often because the email doesn't have a specific claim worth surfacing. That's an email content problem, not a subject line problem.
Real Subject Line Performance Data from 2025-2026
A B2B software company I advised shifted their newsletter subject line format in October 2025. Before the shift, they used curiosity-style subject lines: "Are you making this mistake with your onboarding flow?" After the shift, they moved to specific report-style subject lines: "5 onboarding drop-off patterns we found in 200 SaaS audits."
The impact over the first three months:
Open rate: up from 24.1% to 27.8%. CTR: up from 2.3% to 3.4%. Unsubscribe rate per email: down from 0.18% to 0.12%. List size: 22,000 subscribers.
The CTR improvement is the more important signal. More people who opened the email clicked through to the content. Specificity in the subject line appears to attract readers who actually want the content, which means higher engagement downstream.
A retail brand on the same platform saw similar results. Moving from "New arrivals you'll love" to "New arrivals: organic cotton basics in 6 new colors" increased open rate by 9% and CTR by 31% over a 60-day test window.
The CTR lift in both cases matters more than the open rate lift. For more on why CTR has become the metric that actually tells you whether your email is working, see how to increase email click-through rate.
Length and Preview Text in the Gemini Era
Subject line length advice has been relatively stable: 40-60 characters performs well across clients because it doesn't get truncated in most mobile views. That hasn't changed.
What has changed is the relationship between subject line and preview text. Preview text (the snippet that appears after the subject line in many clients) used to be a secondary reinforcement tool, mostly useful for adding context that didn't fit in the subject line.
Now, preview text feeds into Gemini's summary generation for some email types. Treating it as dead weight is a mistake. The preview text should extend the specific claim in the subject line, not repeat it and not fill it with filler text. "30% off all linen pieces: ends Sunday April 13" in the subject line, followed by "Shop new arrivals including the Marseille dress, Paloma trousers, and Riviera jacket" in preview text, gives both human and AI readers a complete, specific picture of what's inside.
For a deeper look at how Gemini processes the first 100 characters of email content and how that affects inbox behavior, see front-loading email content for Gemini's 100-character rule.
Sender Reputation Still Matters More Than Subject Line Tricks
Here's the uncomfortable truth that most subject line advice papers over: sender reputation is upstream of subject line performance. If your domain has poor deliverability, a perfectly optimized subject line won't save you. The email won't reach the inbox in the first place.
I've seen teams spend weeks A/B testing subject line formats while their list had a 15% hard bounce rate quietly destroying their sender score. When the bounce rate came down (by using a service like Bulk Mail Verifier to clean the list), their open rates improved by more than any subject line test had ever moved them.
The same logic applies to engagement signals. Gmail's AI is watching how subscribers interact with your emails over time. A history of low engagement, high deletions, and spam reports will depress performance regardless of how well the subject line is written. Strong subject lines are multipliers. They work best on a clean, engaged list with healthy sender reputation.
For more on the full picture of what affects inbox placement, see mastering sender reputation for email inbox placement.
Practical Subject Line Checklist for 2026
Most of the best subject line decisions come down to a short set of questions you ask before you write:
What is the single most specific claim this email makes? Start with that. If you're announcing a sale, what's the discount and on what? If you're sharing a report, what's the most surprising finding? If you're sending a newsletter, what's the one piece of content a reader would most want to read?
Does this subject line give both a human and an AI enough to classify the email type accurately? A promotional email should read like a promotional email. An informational one should read like it contains useful information. Mismatches between subject line tone and email content confuse subscribers and, increasingly, AI classification systems.
Would a journalist use this headline? If the subject line would only work as clickbait and not as a straight news headline, reconsider.
Does the preview text extend the subject line or repeat it? If it repeats, you've wasted the second row of inbox real estate.
Have you removed manipulation signals? Excessive punctuation, all-caps words, and false urgency phrasing can still trigger spam filters and reduce trust signals in Gemini's classification.
Writing Subject Lines for Two Audiences at Once
The mental shift I've found most useful for my own clients is this: you're now writing for two readers simultaneously. The human subscriber who scans their inbox quickly and the AI system that processes your email to decide how to surface it.
These audiences share preferences more than you'd expect. Both respond to specificity. Both are put off by manipulation. Both reward clarity. The curiosity-gap tricks that exploited human psychology don't translate to AI systems, and they're losing their grip on humans too as inboxes get smarter.
Writing for both audiences at once sounds complicated. In practice, it usually means writing cleaner, more specific subject lines than you were writing before. Strip the cleverness. Surface the actual claim. Trust that the people on your list who want what you're offering will open when they know exactly what's inside.
The compounding effect is worth understanding. Programs that consistently produce specific, honest subject lines over three or four months build a sender reputation signal that Gemini weights positively in future sends. The same email from a sender with a history of clear, accurate subject lines gets more prominent placement than the same email from a sender whose subject lines have historically misled subscribers. This is the part of the shift that most articles on subject line tactics miss: every subject line you send is training data for how Gmail treats your future sends. The quality of your subject lines this quarter affects the placement of your subject lines next quarter. For a broader view of how this reputation-building works across the whole Gemini-era inbox, see the Gmail Gemini overview for marketers.
Tomorrow, pull your last 10 email subject lines and score them against one question: "Does this tell a subscriber specifically what they'll get if they open?" For every one that doesn't, rewrite it with the most specific claim the email actually contains. Then look at whether your CTR improves over the next 30 days. The answer will be clearer than any A/B test you could design.