Privacy, AI and the inbox

Two things are changing at once: what you are allowed to measure, and what sits between your message and the reader.

Open rate is finished as a metric

Apple Mail Privacy Protection pre-fetches images regardless of whether the recipient looked, which marks messages as opened. Given Apple Mail's share, a large and unknowable portion of your open rate is fictional.

The consequences are not cosmetic:

• Open-rate A/B tests measure noise. Subject-line testing on opens is over.
• Engagement segmentation on opens is unreliable. Use clicks.
• Historical comparisons across the change are invalid.

Anyone still reporting open rate as a headline number is reporting a number they cannot defend. Clicks, replies, conversions and unsubscribes are what remain.

Sender requirements became rules

Gmail and Yahoo moved from recommending to requiring: authentication for bulk senders, one-click unsubscribe honoured quickly, and spam complaints kept under a hard threshold.

The direction is clear and unlikely to reverse — expect thresholds to tighten and expect other providers to follow. Treat the requirements as a floor rather than a target.

AI in the inbox

Mailbox providers and clients are adding summarisation, categorisation and prioritisation. The reader increasingly encounters a machine-generated impression of your email before the email itself.

Early, and worth watching rather than reorganising around. What is safely true: an email with a clear single point summarises well, and one that buries its purpose in preamble does not. That is the same advice as always, with a new reason.

Privacy law keeps expanding

Consent requirements continue to spread beyond Europe, with more jurisdictions adding response-time obligations for access and deletion requests and meaningful revenue-based penalties. If you hold subscriber data across markets, the strictest applicable rule is the practical standard.

The strategic read

Everything here pushes the same way: toward smaller, better-consented, more engaged lists and away from volume. The channel is becoming more like the rest of marketing — measured on outcomes rather than proxies, with permission as the scarce input.

What trips people up

• Reporting open rate to stakeholders who will eventually ask what it means.
• Treating sender requirements as advisory.
• Optimising for a metric that a privacy feature can invalidate overnight.

You have got this when

Your reporting no longer leads with open rate, and you could pass Gmail's bulk sender requirements today without changing anything.