AI email subject lines: how to improve open rates

You send the same newsletter to the same list every fortnight, and the open rate keeps sliding while nobody on the team can say why. Rewriting subject lines by hand takes an hour a send and the number barely moves. AI email subject lines and send-time tools are now built into Klaviyo, Mailchimp and ActiveCampaign, and they can take that hour off your plate. Before you switch them on, it helps to know what each tool actually does, which plan it sits on, and why the open rate you are trying to lift has become an unreliable score.

Most of the opens your platform reports now come from Apple devices

In July 2026, Apple's mail clients accounted for 62.26 percent of the more than one billion opens in Litmus's email client market share report. Since iOS 15, opens from people who switched on Apple's Mail Privacy Protection stop passing along details of the client and platform, and Litmus says those opens are not considered reliable.

Klaviyo handles this with a flag. Every open event carries an Apple Privacy Open property set to true or false, and Klaviyo's help centre says it has no way of telling a person's open from an automated one when Mail Privacy Protection is on. That changes how the rest of this guide reads. A subject line test scored on opens is partly scored by machines, so judge subject lines on clicks and placed orders, and build your engaged segments with Apple privacy opens filtered out.

What personalised subject lines and better timing can change

Personalisation still earns its place. Campaign Monitor states that readers are 26% more likely to open emails with personalised subject lines, though the page does not publish the study behind the figure. Klaviyo's product page claims an average 63x return on email, and says automated emails drive more than 14x higher revenue per recipient than manual campaigns. Both are vendor figures, and both suggest the channel still pays when the message and the timing are right.

The Litmus State of Email 2026 report found that advanced AI adopters are 75 percent more likely to report email ROI above 45:1. That comes from a survey of marketers describing their own programmes, so treat it as a link between AI adoption and strong email programmes, without assuming the AI produced the return.

What the AI writes, and what it learns from

The subject line tools are simpler than the marketing around them suggests. In Klaviyo, you describe the send in the subject line field and the subject line assistant suggests three options, which you can pick, edit or regenerate, in both campaigns and flows. You still decide what the campaign says, and the assistant drafts wording around your description. Mailchimp's generative features come from Intuit Assist, which Mailchimp announced in September 2023 as a way to draft a campaign from subject line to call to action.

Send-time tools learn from behaviour, and the behaviour they learn from matters because of the Apple problem above. Mailchimp's Send Time Optimization picks the time within 24 hours of your chosen date when contacts are most likely to open. Klaviyo's Personalized Send Time looks at opens, clicks and placed orders, then sends the same campaign to each recipient at a different hour inside a window you set. ActiveCampaign's predictive sending times each email around a contact's habitual open times. A model that learns mostly from opens will pick up some of Apple's timing along with your readers' timing, which is one more reason to judge the result on clicks and orders.

We checked every row against the vendor's own help pages on 2 October 2026. Plan names and limits change, so open the page again before you buy an upgrade for one feature.

Set it up on the list you already have

Audit the last 90 days on clicks and orders

Pull the last 90 days of campaign data and look at click rate and placed orders by day of week, time of day and segment. In Klaviyo, rebuild any engaged segment with the condition that Apple Privacy Open equals false, so the segment reflects behaviour you can trust. If one segment clicks at lunchtime and another in the evening, a single send time is serving one of them badly. For a complete walkthrough of automated sequences, see our guide to AI email nurture workflows.

Let the assistant draft, and keep the brief yours

Write a one-line description of the send that covers the offer, the audience and the tone, and let the assistant draft from it. Keep two or three of the suggestions, A/B test them on a portion of the list, and pick the winner on clicks. If your platform has no built-in assistant, a Make or Zapier scenario can pass the draft subject line to ChatGPT or Claude and write the rewrite back before the send. We walk through that pattern in our guide to email personalisation at scale with ChatGPT and Mailchimp.

Switch on send-time optimisation for campaigns without a deadline

Klaviyo recommends Personalized Send Time for non-urgent campaigns, where reaching each person at a good hour matters more than everyone getting the message at once. Flash sales and anything with a deadline should keep a fixed time. In its launch post, Klaviyo says the top-performing campaigns in its beta saw a 35 percent increase in click rates, and adds that results may vary by implementation, industry and business model. Every campaign sent this way is tested against a control group sent at a set time, which gives you a clean comparison inside your own account.

Measure the lift in clicks and revenue

Compare the next four weeks with the previous four on click rate, placed order rate, unsubscribe rate and revenue per recipient. For the business case, take last month's campaign revenue and divide it by the clicks that produced it to get your revenue per click. Multiply that by the extra clicks per campaign your control-group comparison shows, then by the number of campaigns you send in a year. If the result is bigger than the cost of the plan that unlocks the feature, the case is made. If the extra clicks are close to zero, keep your fixed send time and put the effort into the message.

Falling engagement can also be an early sign that a customer is drifting away. Our guide to AI churn prediction shows how to flag those customers before they stop reading altogether.

Questions about AI subject lines and send times

This is the kind of setup we build at Supernodes. Speak with us if you want the audit run against your own list data.