For twenty years, email marketing assumed a person opened a message and read it. That assumption is eroding. Gmail's Gemini and Apple Intelligence now summarize, prioritize, and sometimes rewrite what subscribers see before they open anything. Brands that adapt can gain ground. Brands that keep optimizing for open rates will be measuring the wrong thing.
The short version: AI summaries reward clear offers and clean sending practices, punish bloated or misleading emails, and make open rate even less trustworthy. Shift your focus to clicks, conversions, revenue per recipient, and genuine engagement.
What has changed
| Platform | What it does | Where it appears | Status |
|---|---|---|---|
| Gmail (Gemini) | AI Overviews summarize emails and threads; Promotions tab sorts by relevance instead of date | Summary card above the message; inbox ordering | Rolled out in January 2026 for thread summaries; wider promotional coverage is gradual |
| Gmail AI Inbox | Re-ranks messages around inferred VIPs and time-sensitive items | Separate inbox view | Trusted-tester rollout, expanding through 2026 |
| Apple Intelligence | Replaces the preview line with an AI-generated summary | Inbox list view on supported devices | Live on supported iPhones, iPads, and Macs with the feature enabled |
| Natural-language inbox search | Answers questions across your inbox | Search | Paid Google AI tiers; expanding |
Exact coverage differs by device, region, and plan, and it is changing. Treat any single stat you read about summary coverage as a snapshot.
What this means for ecommerce email
1. Open rate is less reliable than ever
Apple Mail Privacy Protection already inflated opens by pre-loading tracking pixels. Reports also suggest Gmail may auto-open messages to generate summaries, which some platforms record as human opens. If your reported open rate is flat or rising while clicks and revenue fall, this is a likely culprit.
What to do: treat open rate as a rough directional signal only. Anchor reporting on click rate, placed order rate, and revenue per recipient. Segment engaged audiences by clicks and purchases, not opens.
2. The summary becomes your first impression
In the AI summary, a model extracts your offer, deadline, and call to action. If your email buries them, the summary will too.
What to do:
- Put the offer, product, and deadline in the first lines of the email
- Use plain, specific subject lines that match the email content
- Keep a single clear call to action
- Write a preheader that complements the subject line, but do not rely on it, since Apple's summary can replace it
- Test your emails by pasting them into an AI summarizer and seeing what comes back
3. Relevance ranking rewards engagement
Gmail has moved the Promotions tab toward relevance-based sorting and is testing an AI Inbox that re-ranks around priority senders and time-sensitive messages. Brands whose subscribers engage get surfaced. Brands that blast an unengaged list slide down.
What to do:
- Send fewer emails to unengaged subscribers, and suppress them from promotions
- Run a sunset policy that stops sending to people who have not clicked or purchased in a defined window
- Prioritize flows over blasts, because behavior-triggered emails tend to be more relevant
- Encourage replies, since reply history is a strong signal of a real relationship
4. Deliverability fundamentals matter more
AI filtering sits on top of the existing sender reputation system. Complaint rates, authentication, list hygiene, and one-click unsubscribe still decide whether you reach the inbox at all. See our deliverability guide for the technical checklist.
5. Visual identity still counts
Expect more emphasis on authenticated brand identity, such as BIMI logos. A verified, recognizable sender helps both human readers and AI ranking.
Old playbook vs AI-inbox playbook
| Area | Old approach | AI-inbox approach |
|---|---|---|
| Success metric | Open rate | Click rate, placed order rate, revenue per recipient |
| Subject line | Curiosity and teasers | Specific and honest; matches the body |
| Email structure | Story first, offer later | Offer and deadline up front, story after |
| Preheader | Primary hook | Supporting line; not relied upon |
| List strategy | Send to everyone | Engagement-based segmentation and sunsetting |
| Cadence | Volume-driven | Relevance-driven, flow-first |
| Testing | Subject line A/B on opens | Test on clicks and revenue; check summaries |
What not to do
- Do not stuff hidden text or tricks to game summaries. Models read the content, and it risks spam filtering.
- Do not write misleading subject lines. A summary exposes the mismatch.
- Do not cut email length just because summaries exist. Clarity matters more than brevity.
- Do not abandon email. The channel is changing, not dying.
How AI affects the rest of your lifecycle program
If the inbox is becoming a layer of AI triage, your owned channels should work together:
- SMS and RCS: a direct line that is not filtered by an email ranking model. Use it for time-sensitive messages.
- Direct mail: a physical touch that no inbox model can summarize away.
- Onsite and loyalty: engagement you control end to end.
- First-party data: zero-party data from quizzes and preference centers lets you personalize without relying on inbox signals.
A 30-day action plan
- Audit reporting. Remove open rate as a primary KPI in dashboards and flow reviews.
- Rewrite your top five flows (welcome, abandoned cart, browse abandonment, post-purchase, winback) so the offer and CTA lead.
- Run a summary test. Paste your last 10 campaigns into an AI summarizer and note what it extracts.
- Tighten list health. Define an unengaged segment and reduce its send frequency.
- Check authentication. Confirm SPF, DKIM, DMARC, and one-click unsubscribe; consider BIMI.
- Rebase your benchmarks on click and revenue metrics, so year-over-year comparisons make sense.
FAQ
Will AI summaries kill email marketing?
No. Email still drives owned-channel revenue. The summary changes how people decide to engage, and rewards clear, relevant messages.
Is open rate dead?
It is unreliable, not useless. Use it as one weak signal among stronger ones.
Does Gmail summarize promotional emails?
Coverage has been gradual and varies. Plan as if it will become standard.
Should I stop using preheader text?
No. Gmail and others still show it for many recipients, but do not make it your only hook.
How does this affect Klaviyo flows?
The same principles apply. Klaviyo's AI tools can help you rewrite flows, but a human should review that the offer and CTA lead and that the copy is accurate.
How ATTN helps
ATTN audits email programs for the AI-inbox era, from deliverability and list health to flow structure and reporting. If your open rates look healthy but revenue per recipient does not, talk to our retention team about an audit.