The operational decisions that separate service from manipulation
My last post ended with the claim that AI is a neutral tool, and that its impact on the customer journey depends on how it’s used. That sounds good in principle, but using it properly isn’t simple. The difference between AI that serves and AI that exploits lies in small operational decisions teams can make without realizing the harm, even with good intentions.
Puntoni and colleagues, writing in the Journal of Marketing, describe the psychological tension at the center of every AI-powered interaction. Customers move between feeling understood and misunderstood, between feeling served and exploited. The same personalization engine can produce either response. What determines the outcome is whether the design respects autonomy or works against it.
To avoid a hit to customer trust, marketing teams need to be specific about what building trust with AI looks like in practice.
Where AI-Powered Marketing Crosses Into Manipulation
The same capabilities that make AI valuable also make it dangerous when used recklessly. AI has made personalization possible at a scale once reserved for a company’s highest-value accounts. A well-built system can now anticipate needs, tailor content, and remove friction for every customer at once. AI adoption won’t slow down any time soon because it helps brands connect with customers at a high level and on a large scale.
Personalization becomes manipulation when that same tool is trained to use what it knows about a customer to exploit a vulnerability rather than meet a need. An algorithm that surfaces a relevant product at the right moment is serving the customer. An algorithm that detects financial stress and times a high-pressure offer to it is exploiting them. While the mechanism is identical, the intent is not, and that’s the distinction.
Dynamic pricing crosses into manipulation when it shifts from reflecting real conditions like demand or inventory to extracting the maximum a specific person will tolerate based on browsing behavior, device, or location. Automated urgency crosses it when the countdown timer resets on every page reload. AI-generated copy crosses the line when it uses personalization to make a manufactured claim feel individually true. In each case, the customer is treated as a target to be optimized against rather than a person to be served.
Not only does this backfire, but there’s research to prove why. Puntoni’s team found that when customers sense they’re being exploited by an AI system, they respond with adversarial behavior. They’ll feed the algorithm false data, disable inputs, and abandon the relationship altogether. At that point, the manipulation has taken hold, and trust has already been lost.
None of this means that those behind the algorithms are bad people. It means the team focused on conversion is focused on the metrics and may not see the cost. SOPs may be needed to help them work through such issues. For now, AI accelerates the loss of trust by removing the friction that once slowed teams down. A manipulative tactic that once took a week to build and test can now be deployed and scaled in an afternoon, which means the guardrails have to be intentional. No one trips into transparency by accident anymore.
What Transparent AI-Powered Design Looks Like
Rather than vilifying AI and abandoning it, we can use those same capabilities toward an objective that prioritizes customer trust as a key metric.
The clearest way to see that difference is to watch how the same service, built two different ways, would treat the same customer. Consider a hypothetical postpartum subscription service called Held, built for first-time mothers in the first ninety days after birth. This is one of the most cognitively overloaded times in a consumer’s life, which makes it exactly the kind of category where manipulation is easy, and trust is fragile. A new mother is exhausted, anxious about doing things correctly, and short on the bandwidth to scrutinize the interfaces she’s moving through. Everything about her situation makes her easy to exploit, which is precisely why the design decisions matter so much.
A manipulative version of Held would use everything it knows about her to drive conversions. Features would include fear-based messaging about whether she’s doing enough, countdown timers on the sign-up page, and a cancellation flow buried three menus deep so that retention numbers look healthier than they are.
A transparent version uses the same data and the same AI capabilities toward the opposite end. Features include personalization that times content to the mother’s recovery stage, which can be hard to find for new mothers, and genuinely helpful. A conversion page would involve a clear path to sign up, and just as clear a path to cancellation without manufactured scarcity. Signing up and leaving would both require the same level of effort. If urgency is used, it’s reserved for information that’s genuinely time-sensitive, like a shipping cutoff, communicated plainly. An email sequence is used for people who visited without subscribing, delivering useful content without pushing sales. It includes an easy opt-down to a content-only newsletter, rather than forcing a choice between a full subscription and nothing.
Each of these is an operational decision that earns a customer’s trust. The underlying technology is identical to the manipulative version, but the intent is different. This version requires the discipline to keep it transparent, even when conversion pressure builds.
An Operational Test for Your Own Ecosystem
Here’s a way to audit whether your AI-powered marketing is building customer trust or depleting it. For each automated touchpoint, ask whether the customer would still be comfortable with it if they could see exactly how it worked. If your personalization logic, your pricing rules, and your urgency triggers were fully visible to the person on the other end, would they feel served or exploited?
A tactic that only works when the customer doesn’t understand it is a tactic operating against them. One that works just as well when fully explained is one aligned with their interests. Any team can write that into their AI-use guides and SOPs, and it can also hold up under conversion pressure.
Some marketing leaders decide early that customer trust is a metric worth safeguarding. They build the operational discipline to hold that boundary even when quarterly pressure makes the manipulative shortcut tempting. Those are the leaders who will get AI right from the start.
