Newsletter subject lines that earn the click

Apple's Mail Privacy Protection has been auto-fetching tracking pixels since September 2021, and Apple devices accounted for about 52% of email opens when Litmus last measured the market that year. A meaningful share of your "opens" are machines fetching images, not people reading. So the honest way to think about newsletter subject lines is this: you are not optimizing the open rate. You are optimizing the click that follows it.

Subject lines are one piece of the craft; the full path from zero is in our guide to starting a newsletter on your own domain, and the writing that comes after the open is covered in how to write a newsletter.

Opens are inflated, so optimize for the click

MPP downloads images, including tracking pixels, when mail arrives on an Apple device, whether or not the recipient ever reads it. At launch, Litmus projected Apple Mail open rates would drift toward 75% as adoption peaked. Five years on, an "open" from an Apple reader tells you the mail arrived. That is all.

This changes what a subject line is for. Pre-2021, the industry graded subject lines on opens, which at least measured curiosity. Now the only numbers that survive contact with MPP are clicks and replies, so a subject line succeeds when the reader who opens goes on to do something. A line that tricks a person into opening and gives them nothing does not just waste the nine seconds Litmus measured; it teaches them to skip you next week. More on which metrics still mean anything in our piece on open rates.

The 40-character budget

Mobile clients cut subject lines short, and the cut points are tighter than most people write to. Email Tool Tester's client-by-client measurements (checked August 2026) put the visible portrait-mode budget at:

The math that falls out of this is simple. Write to 40 characters or fewer, put the word that carries the meaning inside the first 30, and let the preheader carry everything that did not fit, since inboxes render subject and preheader as one continuous line. "New: one-click rollback for failed deploys" runs 42 characters and loses its last word in the Gmail app; "One-click rollback is live" keeps the payload inside every budget and leaves "for failed deploys, no redeploy needed" to the preheader.

People ask us for catchy newsletter subject lines. Catchy is the wrong target at 40 characters. Clear beats catchy in every inbox we have ever read, including our own.

Newsletter headline examples by type

The sets below are ours, written for this article for three imaginary senders: a coffee roaster, a local city brief, and a developer tools changelog. We have not A/B tested them at scale, and we would distrust anyone who claims performance numbers for a list of unsent subject lines. Character counts are in parentheses. (Inside the email, the same patterns work as section headlines; the constraint that changes is the 40-character budget, which only applies in the inbox.)

Curiosity lines open a loop the issue must close:

• The blend we pulled after 12 days (33)
• What the inspector found upstairs (33)
• We deleted 4,000 lines of config (32)
• The feature nobody asked for (28)
• Our worst issue, by the numbers (31)

Curiosity works when the payoff is real and arrives early in the issue. Its failure mode is fatal, though: tease without a cashout twice and readers stop believing you, which is a worse outcome than never being opened.

Specific lines state the payload outright:

• Ethiopia Guji is back Thursday, 250g bags (41)
• Saturday: the harbor market moves indoors (41)
• One-click rollback is live (26)
• Free shipping ends Sunday night (31)

Specific is the default we recommend. It self-selects the right readers, and the click it earns is a qualified one. The tradeoff: a specific line gets skipped by everyone the payload does not apply to, and you have to be at peace with that.

Number-led lines promise a countable read:

• 3 breweries, 1 permit fight, 12 new jobs (40)
• Our January numbers: 41% fewer tickets (38)
• 5 links on caching, 2 worth your weekend (40)

Numbers signal that the issue respects the reader's time. They fail when every issue leads with a number, because the pattern reads as a template by week four. Rotate.

The spam-words reality check

The lists of "spam trigger words" that circulate every year are mostly folklore. Our stance, and the stance of the providers' own documentation: mailbox filters decide folders on sender reputation, authentication, and complaint rates, and vocabulary is a minor signal at the margin. Google's bulk-sender requirements (February 2024) demand SPF, DKIM, DMARC, one-click unsubscribe, and a spam-complaint rate under 0.3%. They contain no list of banned words. Even Mailjet, which publishes a spam-words guide, concedes in it that a sender with a clean reputation can say "free" without incident. Note the example set above: "Free shipping ends Sunday night" is a fine subject line on an authenticated domain with a warm history.

We have first-hand evidence that the folder decision lives below the content layer. In June 2026, Gmail silently rejected 73% of the mail we were forwarding between our own systems. The cause was DMARC alignment broken by forwarding. Not one word of any subject line was involved. If your mail is landing in spam, audit your authentication and your list before you audit your adjectives; the full diagnostic order is in why emails go to spam.

A/B testing under 2,000 subscribers is noise

Subject-line split tests need more sample than most newsletters have. The arithmetic: to detect a 5-point lift on a 40% baseline (40% vs 45%) at the standard 95% confidence and 80% power, you need roughly 1,500 recipients per arm, more than 3,000 subscribers split down the middle. Run that same test with 1,000 per arm and the smallest difference you can reliably detect is about 6 points, a swing real subject lines rarely produce. You can rerun this in any two-proportion sample-size calculator; Evan Miller's free one is the standard. And that math is on opens, the metric MPP already corrupted. Click-based tests need even more sample, because clicks are rarer.

So under about 2,000 subscribers, a per-issue split test is a random-number generator with a dashboard. Test across issues instead: pick one variable (say, specific vs curiosity), hold your send day and voice constant, run each style for a month, and compare clicks and replies over the whole period. Slower, honest, and it measures the thing that matters.

What we track instead

Two numbers: clicks and replies. Mailbox providers weight engagement in roughly that order themselves, replies above clicks above opens, so the metrics that tell you the truth are also the ones that build your sending reputation. Keep the From name and address identical on every send while you evaluate anything, because a changed sender identity resets the reader's pattern recognition and contaminates whatever you were measuring.

Write the subject and preheader together, spend the first 30 characters on the operative word, and grade the result on clicks. Opens flatter you. Clicks pay you.