Email remains one of the most essential levers in digital marketing. But since AI has made its way into email inboxes, the way a message is read, sorted and measured has changed. Here’s what this means for advertisers, and how to adapt your campaigns.
AI moves into the inbox
More and more often, the first reader of an email is an artificial intelligence. In January 2026, Google announced that Gemini was coming to Gmail, bringing automatic conversation summaries, suggested replies and, in testing, an AI-organized inbox. These features are rolling out first in the United States and in English, before reaching other countries and languages.
Gmail isn’t alone. Apple Mail and Yahoo Mail also offer AI-generated summaries. In Apple Mail, these summaries can even replace the first lines of the message in the inbox list. In practice, recipients no longer always discover an email through its subject line and preheader, but through what an AI has taken from it.
These tools read text, though, and pay much less attention to images. An email built around one large visual, where the offer and its terms exist only as an image, risks being poorly summarized, or even flagged as low priority.
Writing emails that both AI and recipients can read
The good news is that the best practices that help AI also help the reader. The first is to write your offer in text, not just in an image. The headline, the discount and the end date should all appear in the body of the message, and every visual should have alt text.
Next, the subject line and preheader should accurately reflect the content. A catchy subject line that has nothing to do with the offer can disappoint the reader, and it can also muddle the summary the AI produces.
Finally, one main message, clear from the very first lines, works better than several offers competing within the same email.
A few practical habits
- The no-images test. View your email with images turned off. If the offer no longer makes sense, the AI won’t understand it either.
- The first sentence as a summary. Write it as if it were the only one read: what the offer is, who it’s for, and when it ends.
- Dates and amounts in plain text. “20% off until December 1” is better than a discount and a date buried in a visual.
- Alt text that describes. “Stand mixer, 20% off” tells the AI something; “header_visual” tells it nothing.
- A specific CTA. “Get 20% off” is clearer than “Discover”, for both the reader and the AI.
- Subject line, preheader and first sentence read together. If they tell the same story, so will the summary.
Open rate is no longer a reliable benchmark
AI is also changing how campaigns are measured. For a long time, the open rate served as the first indicator of a send’s performance. Today it’s losing its value: when an AI scans a message to summarize or sort it, it can trigger an open without any human having read it.
This isn’t entirely new. Since 2021, Apple Mail has offered a privacy protection feature that automatically loads email content, including the pixel used to track opens. For users who have turned it on, an email is counted as opened whether it was read or not.
So an email can show strong open rates without having been seen, and conversely be read in a summary without being counted as opened. To steer a campaign, it’s better to rely on what recipients actually do: clicks, sales and, more broadly, engagement over time.
In practice, compare your campaigns on click-through and conversion rates rather than opens. To find out what works, test two versions of the same email with comparable audiences, and keep the one that generates the most clicks and sales.
Engagement over volume
With AI, an email now has two readers: the machine that sorts and summarizes it, and the person who decides whether to click. Both expect the same thing: a clear offer, written in text, and a message that delivers on its promise. What matters is real recipient engagement, far more than the volume sent.