What changed in the inbox, and what to do about it

Apple Intelligence generates its own preview text in some views, replacing the preheader you wrote. Add the newer inbox categories on top and you control less of that first impression than at any point in the last decade.

You cannot opt out of this, so the sensible response is to stop treating the preheader as a punchline that completes the subject line. If a machine can bin it, the email should still make sense without it.

What I do about it

Subject lines that stand alone. Preheaders that add a second concrete detail rather than the payoff to a joke. And a top-of-email opening that reads well as a summary, because sometimes that is exactly what it becomes. More on writing structure in the ecommerce email guide.

Gmail's Manage subscriptions view lists the senders a person can unsubscribe from, sorted by volume, heaviest first. If you email daily, you are near the top of that list, in front of everyone who has ever half-considered leaving.

The interface itself now penalises frequency without value. That is a real change: it used to be that over-sending cost you engagement slowly. Now there is a screen that hands people a one-tap exit and puts the noisiest senders first.

It is not an argument for sending less on principle. It is an argument for every send being worth the slot. If you cannot say what someone gets from an email, it is a liability, not a campaign.

What I do about it

A sending rhythm the shop can sustain, with a mix people benefit from rather than eight promotions a month. The rhythm question is covered properly in the small business guide.

This one is my own verdict rather than a product announcement, but it follows from the rest. Apple's Mail Privacy Protection pre-loads images and inflates opens. AI pre-scanning fires tracking pixels with no human involved. Grouped and previewed interactions never register at all.

So the open rate is now a number that moves for reasons unrelated to whether anyone read anything. It is still fine as a rough directional signal on a single audience over time. It is useless as a headline metric and dangerous in a report.

What to report instead

Revenue per recipient. Click-to-conversion. Flow revenue split by individual email. Repeat purchase rate from email. List growth net of unsubscribes. Every one of those is downstream of a person doing something deliberate.

The reporting setup I use each month is described on the pricing page, and the platform-specific attribution traps are in the Shopify guide.

The old Postmaster Tools High, Medium, Low reputation grades for domain and IP went with v1. A lot of people still ask for that score and it no longer exists.

v2 gives you a Compliance Status checklist against Gmail's sender requirements, and since June a Deliverability Analysis verdict, which is closer to a straight answer about whether Gmail users want your mail.

It is a better tool and a worse comfort blanket. There is no single grade to screenshot, so checking it has to become a habit rather than something you do when things feel wrong.

What I do about it

Postmaster Tools verified on the sending domain, checked monthly alongside spam complaint rate and authentication status, and reported in plain words rather than a colour.

Average spam placement across sends currently runs anywhere from zero to about 15%. For a lot of senders that means a real slice of every campaign never reaches a person.

The problem is that your email platform cannot see it. Delivered means accepted by the receiving server, not placed in an inbox. A campaign can be 98% delivered and 12% in spam at the same time, and the dashboard will show you a clean green number.

So placement has to be measured separately, with seed testing and by watching for the tells: one mailbox provider performing far worse than the others, click rates dropping while list size grows, or a sudden fall in revenue with no change to the creative.

What I do about it

Authentication set up properly, complaint rate watched against Gmail's threshold, inactive segments suppressed, and placement checked rather than assumed. The audit covers all four on first contact.