Your AI app works. So why do users reject it?
Picture a product manager: sharp, busy, not remotely a technophobe. She uses AI to summarize her meetings, draft her emails, clean her spreadsheets. Then she opens an AI tool to build the deck for the biggest pitch of her quarter. It generates a perfectly reasonable presentation in twelve seconds. The structure is fine. The copy is fine. And she closes it, opens a blank file, and starts over by hand.
Nothing was wrong with what the machine made. So why did she throw it away?
This is one of the most misread moments in the entire conversation about AI adoption. We tend to explain resistance with two lazy stories: either people don’t understand the technology yet, or they’re afraid it will take their jobs. Both are sometimes true. But research in consumer psychology points to something far more interesting, and far more useful if you build or market anything. People don’t reject AI because it’s bad at the task. They reject it because of what accepting it would say about them.
Once you see that, the pushback stops looking irrational. And, more importantly, it stops looking permanent.
“It doesn’t know I’m different”
Start with the most human objection of all: this thing is treating me like everybody else.
In a landmark set of nine studies, marketing researchers Chiara Longoni, Andrea Bonezzi, and Carey Morewedge found that people were consistently reluctant to be treated by “medical AI”, even when they were told the AI was as accurate as, or more accurate than, a human doctor (Longoni, Bonezzi & Morewedge, 2019). That last part is the twist. This wasn’t about competence. People turned down the better option.
The reason they named is beautifully simple. They called it uniqueness neglect: the gut sense that an algorithm sees you as a statistical average, a row in a spreadsheet, while a human sees you as you, your specific body, your specific situation, your specific story. Tellingly, the effect was strongest in the people who thought of themselves as most unique. The more special you feel, the more an algorithm’s one-size-fits-all logic feels like an insult.
Here’s the part every marketer should tattoo on the wall: the resistance had almost nothing to do with the machine’s actual behavior and everything to do with the meaning the person assigned to being served by a machine. You are not fighting a capability gap. You are fighting a feeling of being flattened.
When the task is really about who you are
Now push the idea one step further. Some decisions aren’t just personal; they’re identity. They’re how we tell the world, and ourselves, who we are.
Eugina Leung, Gabriele Paolacci, and Stefano Puntoni ran six studies on exactly this and found that automation becomes undesirable the moment identity motives enter the room (Leung, Paolacci & Puntoni, 2018). Their examples are wonderfully ordinary: cooking, fishing, driving. A person who deeply identifies as “a cook” doesn’t want the machine to make the meal, because then they didn’t make the meal, and the whole point was to be the kind of person who makes the meal. Automation is welcome when it does something for you. It’s resented when it does something instead of you, and that something was part of your self-image.
This is why AI feels totally fine for reconciling an invoice and totally wrong for writing your wedding speech. Same underlying technology. Completely different relationship to your sense of self.
When doing it yourself is the entire point
There’s a third layer, and it’s the one businesses understand least. Sometimes the effort is the reward.
Emanuel de Bellis and colleagues call this the meaning of manual labor, the meaning people draw from doing a task with their own hands (de Bellis, Johar & Poletti, 2023). Their research shows that fully autonomous products, the ones that promise to remove you from the loop entirely, run straight into this wall. A bread machine that does everything can be less appealing than one that leaves you a step to perform, because the person didn’t want bread delivered; they wanted to have made bread. Newer work extends the same logic from physical tasks to mental ones, the meaning of cognitive labor, which matters enormously for any AI aimed at creative or knowledge work, where the thinking was supposed to be the good part.
Get Alara Akcasiz’s stories in your inbox
Join Medium for free to get updates from this writer.
Stack these together and a pattern emerges: We resist AI when it touches uniqueness, identity, or meaning, and we resist it hardest in the domains that are symbolic, expressive, or hedonic rather than purely practical. It’s why people are cool on AI-designed luxury goods (Xu & Mehta, 2022), why we prefer human hands on the things that are supposed to say something about us (Granulo, Fuchs & Puntoni, 2021), and why we happily let AI pick the most durable dishwasher but bristle when it picks the most beautiful painting. As Longoni and Luca Cian put it in their “word-of-machine” research, we lean toward the machine for functional attributes and away from it for experiential ones (Longoni & Cian, 2022).
The plot twist: People don’t actually hate AI
If the story ended here, the takeaway would be depressing and wrong: keep AI away from anything that matters.
But the same body of research contains the escape hatch, and it’s the most commercially valuable finding of all. People are not anti-AI. In plenty of contexts they actively prefer it. Given a numerical estimate to make, people often trust an algorithm’s judgment over a human’s; researchers literally named this “algorithm appreciation” (Logg, Minson & Moore, 2019). For objective, checkable tasks, the machine is the reassuring choice, not the threatening one (Castelo, Bos & Lehmann, 2019).
So the real world isn’t “humans vs. machines.” It’s a line. On one side sit the utilitarian, objective, get-it-done tasks where people want AI to just handle it. On the other sit the tasks tangled up with uniqueness, identity, and meaning, where people want to stay in the chair. Resistance isn’t a wall around AI. It’s a border between two kinds of work.
And here is the finding that should change how you design and sell anything: that border moves. In the original medical-AI studies, the deep resistance people felt didn’t require a smarter algorithm to dissolve. It dissolved when the AI was framed as personalized to the individual, or when it was positioned as supporting a human decision rather than replacing it (Longoni, Bonezzi & Morewedge, 2019). Change the story from “the machine decides for you” to “the machine helps you decide,” and the very same people stop pulling away.
Read that again, because it’s the whole game. You don’t overcome uniqueness neglect with more automation. You overcome it by putting the person back in the frame.
What this means if you build or market with AI
The practical lesson is almost the opposite of where most AI product marketing goes. The default pitch is “look how much it does without you”, hands off the wheel, zero effort, done in seconds. For anything touching identity, taste, or craft, that pitch is precisely the trigger. It’s uniqueness neglect delivered as a value proposition.
The tools that win the resistant, high-value customer do three quieter things:
They do the labor nobody wanted anyway, the setup, the formatting, the blank-page dread, while leaving the meaningful moves to the human. They keep effort where effort is felt as reward and remove it where it was only friction.
They treat you as specific, not average. The output looks like it was made for this person and this situation, not stamped from the same template as ten thousand others. Personalization isn’t a feature here; it’s the antidote to the core objection.
And they let the human sign the work. The person stays the author. The AI is scaffolding, not the building. What you hand your boss is still yours.
A quick worked example
Consider the deck our product manager threw away. A presentation is a near-perfect storm of all three tripwires. It’s identity (”this is my thinking, my pitch”). It’s craft (the effort of shaping an argument is part of the value). And it’s symbolic, it represents you in a room you care about.
It’s also why so many AI presentation tools quietly disappoint. Point a generic generator at a prompt, and you get a deck that is competent and utterly interchangeable: the same layouts, the same stock rhythm, the same this-was-clearly-made-by-a-machine sheen that every other user gets. That sameness isn’t a cosmetic flaw. It’s uniqueness neglect rendered in pixels. The tool just told your customer, in visual form, you are one of many. No wonder they open a blank file.
The more interesting design move, and, full disclosure, the one we’ve built Decktopus around, is to take the research seriously. Let the AI absorb the labor that carried no meaning: structuring the flow, drafting the scaffolding, handling the layout math that makes people miserable. But produce something that doesn’t wear the mass-produced AI look, so the person isn’t handed the “just another data point” feeling in slide form. And keep them in the author’s chair the whole way, shaping and signing the result. Do the boring parts; protect the parts that felt like you. That’s not a UX preference. It’s the exact seam the research says separates the AI people abandon from the AI people adopt.
The takeaway
The next time a smart, capable person rejects a perfectly good AI output, resist the urge to file it under fear or ignorance. They’re telling you something precise: this touched a part of my life where I need to still be the one doing it.
That’s not a problem to argue them out of. It’s a map. It tells you exactly which labor to take off their hands, and which to leave reverently alone. The companies that read the map will build AI people that are relieved to use. The ones that don’t will keep shipping twelve-second miracles that land in the trash, technically excellent, and quietly insulting.
The winning idea was never “AI instead of you.” It was always “AI, so there’s more of you in the work that counts.”
