An iOS update pushed Apple Intelligence onto writer Polly Hudson’s phone, and the suggested replies it generated fooled her husband entirely.
The episode unfolded after a software update added the feature to Hudson’s device. She tried to disable Apple Intelligence, set Siri back to its previous behavior, and turn off suggested replies for texts and emails, but the settings made no difference, she wrote in the Guardian.
Days later, bored on a bus, she began answering her husband’s messages with the AI’s proposed responses to see if he would notice. He did not.
Even lines like “You’re a classic!” and “Haha I know right” — delivered without a comma — passed without comment.

Then the system offered a reply that ended with the private nickname Hudson uses for her husband, not a generic term like “darling” or “honey.” She described the moment: “A chill descended in my soul as the penny dropped that these weren’t generic phrases everyone gets – they were specific.” The phone was impersonating her, trained on her own messaging history.
Apple Intelligence’s Smart Replies appear in Mail and Messages as a suggestions bar above the keyboard. They are designed for messages with clear questions — “Do you want to go to the movies tonight?” — and are less useful for open-ended conversation, according to MacRumors’ guide to the iOS 18.1 rollout.
The feature requires an iPhone 15 Pro, any iPhone 16 model, or an iPad or Mac with an M-series chip, and it shipped in the iOS 18.1 release at the end of October 2024.
Hudson’s experience highlights a practical gap: the off switches she toggled — disabling Apple Intelligence, reverting Siri, deselecting suggested replies — did not stop the suggestions from appearing. CuriousCats noted that she “tried to disable Apple Intelligence and suggested replies but found the options made no difference.”
The AI’s ability to mimic Hudson’s idiosyncratic phrasing — exclamation marks, “banz” with a z, the private nickname — left her feeling “easily replaced” and critical of her own communication style, which she called cheesy and heavy on exclamation marks.
“Further proof, not that it was needed, that resistance is futile,” she concluded, adding that she had “hate-used” the very system she tried to remove. CuriousCats summarized the episode as a demonstration of how AI-suggested replies can be tailored to an individual’s voice rather than generic.
No independent verification of the husband’s non-reaction exists beyond Hudson’s own account.