There is a category of purchase where your best salesperson is a liability. Not because they are bad at the job, but because they are a person.
A piece in Forbes this week pulled together a growing body of research with the same conclusion. When a purchase or a question carries embarrassment, people prefer to deal with AI.
Most marketing teams have not absorbed what that means for conversion design. Worse, the most popular chatbot advice points in the wrong direction.
What the research found
The strongest evidence comes from Jianna Jin at Notre Dame's Mendoza College of Business, working with colleagues from Ohio State and published in the Journal of Consumer Psychology. Across seven experiments with more than 6,000 participants, shoppers looked for products such as diarrhoea medication, haemorrhoid cream, anti-wrinkle cream and hay fever treatment.
For embarrassing products, the preference was lopsided. More than 80 percent of participants seeking diarrhoea treatment preferred a chatbot pharmacist to a human one, and they were more willing to engage and share personal details with it.
That held even when participants were shopping alone at home. Nobody could see them, the judgement they were avoiding was imagined, and it still changed behaviour.
The mechanism matters for design. Embarrassment is a social emotion, and it needs an audience, even an imagined one. A chatbot that is clearly software removes the audience, so the emotion has nothing to attach to.
The data-sharing finding deserves attention too. Customers who feel unjudged tell you more about their situation, which means better product matching and better qualification data. It also means you are holding more sensitive information than before, and the trust that made people share it disappears the first time it is misused.
Separate research from the University of Kansas found the same preference for discussing embarrassing health information, with one important reversal. When people were angry, they wanted a human.
The human-like chatbot backfires
Here is the finding that should change your next chatbot brief. When the bot looked human, with a profile photo and a natural conversational style, embarrassed customers were more likely to assume a real person was on the other end, and many avoided the conversation entirely.
For three years the default advice has been to make bots warmer, more personal and more human. Give it a name, a face and some personality.
For a large set of purchases, that advice is backwards. The value of the bot is precisely that it is not a person, and every design choice that blurs the line removes the value.
The researchers' recommendation is simple. In sensitive categories, make the chatbot clearly and unmistakably a machine.
Embarrassment is bigger than pharmacy
It would be easy to file this under health retail and move on. That would be a mistake, because the underlying driver is fear of judgement, and fear of judgement shows up everywhere.
The researchers point to car buying, where customers expect to be judged on what they know and what they can afford. The same applies to personal finance and debt, insurance claims and cosmetic procedures.
It applies in B2B too. A marketing director who does not understand how attribution works in GA4 will not ask your sales rep, because asking signals incompetence to a vendor. They will ask an AI assistant privately, and your brand is either part of that answer or it is not.
Price is a version of the same thing. A good share of "contact us for pricing" friction is not about the number. It is about the fear of looking like someone who cannot afford it.
I wrote about a related trust dynamic in why selling AI as a replacement kills trust. The nuance here is that customers do not want AI instead of people everywhere. They want it at specific moments, and they want humans at others.
There is a visibility angle as well. The questions people are too embarrassed to ask a person are exactly the ones they now ask assistants. I covered how Meta put an AI agent in every business chat, and the brands that answer those private questions well inside AI interfaces will win customers who never raise a hand.
Designing for the moment of embarrassment
The practical work is a mapping exercise. List the points in your funnel where a customer might feel judged: a basic question, a budget conversation, a sensitive need, an admission that something went wrong.
At those points, offer an obviously automated option first. Label it plainly, and let people get information without identifying themselves.
Then design the exit. Kansas found that angry people want humans, so complaint, cancellation and escalation paths should reach a person quickly, without forcing the customer back through the bot that helped them buy.
Finally, measure the right thing. Track completion and conversion at those sensitive steps before and after the change, not overall chatbot satisfaction, which will average the effect away.
If you want this built into your customer journey without building it yourself, difrnt.ai (difrnt.ai) builds custom AI solutions end to end.
Sometimes the most considerate thing you can offer a customer is the option not to talk to anyone.