Consumer trust in AI-driven personalisation is moving in the wrong direction, even as more brands invest in it. Salesforce's State of the AI Connected Customer report, covering more than 16,500 consumers and business buyers, found only 42% now trust businesses to use AI ethically, down from 58% just a couple of years earlier, a 16-point drop that most marketing teams have not yet caught up to in how they design campaigns.

What actually makes personalisation feel creepy

CMSWire's 2026 analysis frames the distinction clearly: customers generally welcome contextual relevance, but become uncomfortable specifically when a brand appears to know more than the customer consciously shared, or uses personal information without clear consent or an obvious benefit in return. The discomfort is not about personalisation itself, it is about the gap between what a customer knowingly gave and what the email seems to know anyway.

Where it reliably works

• Behaviour a subscriber consciously performed on your own site or in your own emails: past purchases, clicked links, viewed products.
• Preferences a subscriber explicitly stated, through a profile update or onboarding survey.
• Segment-level personalisation based on clear, expected categories, like industry or product line, rather than inferred personal details.

Where it tends to cross the line

• Referencing information a subscriber never knowingly gave you, particularly anything inferred rather than stated.
• Personalisation around sensitive categories, health, financial hardship, or major life events, unless the subscriber has explicitly opted into that context.
• Dynamic content so specific it reads as surveillance rather than relevance, the 'how did they know that' reaction rather than 'this brand gets me'.

The human-voice counterweight

A 2026 Adobe Express study found 37% of consumers say they trust brands more when marketing emails feel genuinely human, even if that makes them look slightly less polished, and 18% said they had actually unsubscribed from a list because an email felt AI-written. Roughly 17% of users separately report annoyance specifically from over-personalised or 'creepy' emails, according to SQ Magazine's 2026 personalisation research. Together, these point to the same conclusion: AI personalisation works best as a tool a human is clearly still directing, not as a replacement for a recognisable human voice.

A practical test before shipping a personalised send

Before sending anything dynamically personalised, ask whether the subscriber would be able to explain, without effort, why you know the detail being referenced. If the honest answer involves inference, third-party data, or anything the subscriber did not knowingly provide, that is the signal to pull back, regardless of how technically impressive the personalisation is.

Frequently asked questions

Is AI personalisation in email losing consumer trust?

Broad trust in businesses using AI ethically has fallen sharply in recent years, though preference for relevant, well-executed personalisation remains high. The issue is less about AI itself and more about transparency in how it is used.

What makes personalisation feel creepy rather than helpful?

Referencing information the subscriber did not knowingly provide, or personalising around sensitive topics without explicit consent, are the two most consistent triggers identified in current research.

Should I stop using AI personalisation in email?

Not necessarily. Personalisation based on a subscriber's own stated preferences or on-site behaviour continues to perform well and build trust. The risk sits specifically with inferred or overly specific dynamic content.

Does a more human tone reduce the benefit of AI personalisation?

No, the two are not in conflict. AI can support segmentation, timing, and drafting while the final voice and judgement still read as genuinely human, which is the combination current research suggests performs best.

Mail Blaze is an email marketing platform specialising in email engagement. AI features are on every plan, and AI credits are billed separately. They are designed to support a human-directed voice rather than replace it.

Explore Mail Blaze's AI tools, built to support your voice rather than replace it.

P.S. If a recent personalised send got an unusual reaction, good or bad, it is worth a closer look at exactly what triggered it. We are happy to help you dig into it.