AI and email marketing in 2026: what's actually working, what's overhyped, and what should worry you
AI is improving some email metrics in measurable ways — HubSpot's own AI-personalized nurture emails lifted its conversion rate 82%. Production time has collapsed from weeks to days. But the same tools are making everyone's emails sound alike, and AI assistants like Google's CC and ChatGPT Pulse already summarize email for users who opt in. The real question is what happens when optimization alone stops being a differentiator?

70% of marketing teams have deployed AI in production, and 88% still edit what it writes. The results are real but uneven — and there's a threat most teams aren't paying attention to.
Key takeaways
- 70% of marketing teams have deployed AI in production; 88% still need moderate to substantial editing of its output (Knak 2026)
- AI subject lines and send-time optimization have no published average lift, so test them on your own list
- AI-personalized nurture emails lifted HubSpot's own conversion rate 82%
- 78% of teams now produce an email in three days or less; in 2024, 62% took two weeks or more (Litmus)
- Google's CC and ChatGPT Pulse already summarize email for users who opt in
I want to be honest about where I land on AI in email marketing, because most of what I read about it falls into two camps: breathless hype or reflexive skepticism. The reality is messier than either camp admits.
HubSpot's own test found AI-personalized nurture emails lifted conversions 82%. That is one company's result, and the widely quoted lifts for AI subject lines trace to no published study. AI is also collapsing production time so fast that teams are producing more emails without necessarily producing better ones. And there's a developing threat — AI gatekeepers sitting between your emails and the subscriber — that could reshape the entire channel.
This post is part of our 2026 Email Marketing Benchmarks series.
AI adoption in email marketing
How many marketers are using it?
| Metric | Share of teams |
|---|---|
| Have deployed AI in marketing production | 70% |
| Still need moderate to substantial human editing of AI output | 88% |
Source: Knak State of Marketing Production 2026
Knak's survey covers marketing production as a whole, so email is one channel among several. Adoption is wide, but most teams still rewrite what the AI produces. Few have rebuilt their workflow around it.
What AI actually does to email metrics
Performance improvements by application
| AI application | Metric improvement |
|---|---|
| 1:1 personalization (HubSpot's own nurture-flow test) | +82% conversion rate |
| Personalization in general (McKinsey, 2021) | +10-15% revenue, typical |
Revenue impact
McKinsey's 2021 research found that personalization in general most often drives a 10 to 15 percent revenue lift, with company results ranging from 5 to 25 percent.
Subject lines: what 6 billion emails a year show
Jay Schwedelson (founder of SubjectLine.com, host of the marketing podcast Do This, Not That) runs an agency that sends about 6 billion emails a year. At INBOUND 2025 he reported what's working:
| Tactic | Open rate impact |
|---|---|
| Priority alerts in subject line | +28% |
| 3 words or less with no preheader | +25% |
| Downplayed urgency (counterintuitive) | +20%+ |
| Negative emojis such as 💔 or ⚠️ | +17% |
Subject line length sweet spot: 28-50 characters.
Alert and urgency tactics carry a risk. There's a line between relevance and aggression, and AI optimization can cross it if nobody's paying attention.
Production speed: the most dramatic shift
| Litmus State of Email | Share of teams |
|---|---|
| 2024: took two weeks or more to produce an email | 62% |
| 2026: produce an email in three days or less | 78% |
The two surveys measured different things, so compare them with care. AI handles first drafts, image generation, subject line variants and code, and production timelines have shrunk from weeks to days.
The upside: email teams can test more variants and iterate faster. The downside is obvious. More output doesn't mean better output, and I'm already seeing inboxes full of emails that read like they were written by the same machine. Because they were.
Send times: what AI optimization actually shows
AI send-time optimization predicts each subscriber's best receive time from their past engagement. Neither Klaviyo nor Brevo publishes an average lift. Klaviyo's Personalized Send Time holds back a control group on each campaign and reports the lift for your own account, with Apple MPP opens excluded. GetResponse's 2024 report, on 2023 sends, put its per-subscriber Perfect Timing at a 39.46% open rate. Its Time Travel feature, which sends at a fixed hour in each subscriber's local time zone, averaged 45.07%.
Validity's 2026 Email Deliverability Benchmark calls send-time optimization increasingly outdated. Its reasoning is that inboxes such as Gmail's promotions tab rank email by the brands each reader interacts with most.
Aggregate data (when you don't have AI STO)
The MailerLite send-time study covers 2.1 million campaigns sent from December 2024 to November 2025:
| Day (MailerLite 2025) | What MailerLite found |
|---|---|
| Every day | Average open rates sit between about 46.5% and 47.5%, so the day matters little |
| Tuesday and Friday | Highest average click rates, closely followed by Monday |
| Thursday | Most campaigns sent |
| Friday 6 PM | Best single hour of the week: 49.7% average open rate and 8.09% click rate |
Best times:
| Time window | Best for |
|---|---|
| 8-11 AM | Opens on most weekdays |
| 8-9 PM | Clicks on most weekdays |
| Friday 6 PM | Opens and clicks, the only day both peak at the same hour |
Average open rates differ by about one point from day to day. If you have AI send-time optimization, test it against a fixed local-time send before you rely on it.
Email clients: where opens happen
| Client | Share of tracked opens |
|---|---|
| Apple | 76.56% |
| Gmail | 16.10% |
| Outlook | 3.48% |
| Yahoo Mail | 2.35% |
Source: Litmus email client market share, September 2026 (1.5 billion opens)
Apple's share includes Mail Privacy Protection opens, so machine opens inflate it, and Litmus no longer publishes a mobile-versus-desktop split. Subject lines and preheaders still need to work in mobile preview panes. Landing pages need to work on whatever device converts best for your audience.
The AI gatekeeper problem
This is the part that should get more attention than it does.
Google's CC and ChatGPT Pulse
Google's CC, a Google Labs experiment launched in December 2025, emails users who opt in a morning briefing built from their Gmail, Calendar and Drive. ChatGPT Pulse can read Gmail once a user connects it, a connection that is off by default. For those subscribers, an AI summarizes your email in its own words, decides whether to surface it, and shapes how they perceive your brand voice.
If AI agents sit between your email and the subscriber, the thing you've been optimizing — the subject line — becomes less important. What matters instead is sender reputation, relationship quality, and content relevance. Those are the signals AI agents use to decide what's worth surfacing.
The sameness problem
Jeanne Jennings (founder of Email Optimization Shop) puts it well: "Marketers who treat AI as a means to deepen relevance, not a shortcut around fundamentals, will outperform."
When everyone uses the same AI tools to optimize subject lines, the optimization advantage disappears. What's left is whether your brand is actually distinctive and whether subscribers want to hear from you. AI can't manufacture that.
Email that reads like every other AI-generated email gives subscribers one more reason to ignore it.
What to actually do with AI in email
Where the returns are clear
- Subject line testing at scale — let AI generate variants, then test them on your own list
- Personalization engine — 82% conversion lift in HubSpot's own nurture-flow test
- Production acceleration — use the freed-up time for strategy, not more volume
- Predictive engagement scoring — prioritize high-value subscribers
Where to be careful
- Full autopilot — AI drafts need human review. Brand voice and factual accuracy both drift without it.
- Over-optimization — if everyone's subject lines are AI-optimized, nobody's stand out. This is already happening.
- The gatekeeper trend — AI intermediaries summarizing emails before subscribers see them. Prepare now.
- Confusing speed with strategy — producing emails faster is not the same as producing better ones.
Related: Are email open rates dead? | The email metrics that actually matter
Frequently asked questions
How much do AI subject lines improve open rates?
No published study gives an average lift for AI-written subject lines. Jay Schwedelson's data, from an agency that sends about 6 billion emails a year, shows which tactics move opens, such as priority alerts (+28%) and subject lines of three words or less with no preheader (+25%). Sweet spot for length is 28-50 characters.
What is AI send-time optimization?
It analyzes individual subscriber behavior to predict when each person is most likely to engage. Most major ESPs offer it now. Neither Klaviyo nor Brevo publishes an average lift, and GetResponse's 2023 data showed per-subscriber timing opening below fixed local-time sends (39.46% against 45.07%). Test it with a holdout group on your own list.
How much does AI personalization improve conversion?
In HubSpot's own test, using AI to pick each lead's content and write a matching email lifted its nurture-flow conversion rate 82%, opens 30% and clicks 50%. McKinsey's 2021 research puts the typical revenue lift from personalization at 10 to 15 percent.
Will AI replace email marketers?
Not yet. In Litmus's State of Email 2026, 78% of teams produce an email in three days or less, but in Knak's 2026 survey 88% of teams still need moderate to substantial human editing of AI output. The role is shifting from execution to strategy and relationship design.
What is the AI gatekeeper threat?
Google's CC and ChatGPT Pulse summarize email for users who opt in. This could shift what matters from subject line optimization to sender reputation and relevance — the factors AI intermediaries use to decide what's worth showing.
Related reading:
- Abandoned cart email: the ultimate recovery guide
- The Modern Ecommerce Email Stack: Why Two Tools Beat One in 2026 — Why AI-driven personalization belongs in the orchestration layer, not the ESP, and why Klaviyo's May 2026 Anthropic partnership confirms the layering thesis.
Sources
-
Litmus State of Email 2026
-
Jeanne Jennings, emailopshop.com
Part of the 2026 Email Marketing Benchmarks series by Geysera

Co-Founder and CEO
Bob Thordarson is CEO and Co-Founder of Geysera, a serial entrepreneur with 25+ years and five co-founded ventures, including Cequint (acquired by TNS in 2010 for $112.5M) and Consumerware (acquired by ParkerVision). A graduate of the University of Washington and MIT Entrepreneurial Masters Program, based in Seattle, he serves on the boards of DRY Soda Co. and the Entrepreneurs' Organization Seattle chapter. He is an expert in retention marketing email systems and methodology for ecommerce and B2B brands — measured by incremental revenue, not vanity metrics.