Part 3 of 3 — Most non-users aren’t waiting to be converted. They’re telling us something about AI itself — and we need to listen.

The first two parts of this series looked closely at people who already use chatbots. This last one is about the people who don’t — and it turns out they are not a queue of future users who simply haven’t gotten around to it. When you listen to them, what you hear is something broader and harder to pin down: a real, somewhat undirected frustration with the whole world of AI as it stands in 2026. They don’t much trust or like the companies building it, and they don’t much trust the technology itself.

We surveyed 100 U.S. adults who hadn’t used a chatbot in the past week. A sample that small carries a margin of error close to 10 points, so treat the exact figures as broad guidance. But the shape of what they told us is clear, and it’s more interesting than a list of missing features.

We asked the same question two ways

We asked why they hadn’t used a chatbot in two different ways, and the order matters. First, openly, in their own words: was last week an exception, a deliberate decision, or just something you haven’t tried? Then, on a later screen, we gave them a checklist of 25 possible reasons and asked them to tick all that applied. Asking the open question first meant people had a clean, unprompted shot at saying whatever was actually on their minds before any list could nudge them.

Asked cold, the answer is mundane

In their own words, the most common answer by a wide margin had nothing to do with ethics or the environment. It was simply not having a use for the thing. “I did not use this last week because I did not need it.” “I never use chatbots or had a reason to.” “For the most part, traditional Google searches fill my information-seeking needs.” Where deeper objections showed up, they were usually bolted onto a practical one: “I’m technologically literate, and have no need to use LLMs for any of my daily tasks.”

What non-users said in their own words, before seeing any list. n = 101; responses may touch more than one theme.

Hand them a list of options, and the mood changes

Moments later, the same people were shown 25 possible reasons. Now they endorsed a lot of them — an average of 6.2 each — and the ones that floated to the top were overwhelmingly about ethics, companies, and the planet. Environmental impact led at 62%, followed by “AI is overhyped and distracts from real human connection,” distrust of the companies behind these tools, and distrust of what they do with your data.

The same reasons, measured two ways. Environmental concern is the top box people tick — and one of the last they raise on their own.

Here’s the tension. Environmental impact is the number-one reason on the checklist, but barely 7% of people raised it unprompted a few screens earlier. Data and privacy: ticked by half, volunteered by almost no one. Meanwhile, “I just don’t have a use for it,” the runaway leader when people spoke for themselves, has no real equivalent near the top of the list.

What the frustration is about

Because we asked the open question first, we can’t wave this away as survey fatigue. People had a clear chance to name any reason they wanted, and most named a practical one. Only when handed a menu of articulate, principled-sounding objections did they start nodding along in large numbers. That tells us the checklist isn’t measuring what drove last week’s behavior so much as what people will agree with when it’s offered.

Our read is that the environmental concern, while certainly real, is doing a specific additional job. While the environmental concern is very genuine — people really do worry about the energy and water that AI consumes and the real impacts of noise pollution — the fact that this concern was expressed so much more in the checklist of choices and not in the first, open-ended question suggests strongly that more is going on here. The environmental concern is legitimate and frankly laudable, but it’s also the most respectable and most defensible way to voice a much bigger and blurrier feeling. “I don’t trust these companies and I don’t like where this is all going” is hard to fit in a survey box. “I’m worried about the environmental impact” is easy, principled, and admirable — so it soaks up the weight of more nuanced concerns. The numbers are so different precisely because “the environment” is standing in for something larger: a broad distrust of the major companies building AI, and of the technology itself.

Environmental concern is the top box people tick (62%) — yet barely 7% raise it unprompted. It’s less the reason than the acceptable language for a deeper distrust.

Why we need to understand this — not argue with it

Understanding this frustration matters more than overcoming it. We can’t and shouldn’t argue someone into a need they don’t have, and you can’t dissolve a diffuse distrust of an entire industry with a feature or a press release. The frustration is real, it’s widely shared, and it’s aimed at the companies and the category as much as at any product. 

Taking it seriously means a few things at once: accepting that for many people the honest answer is “I don’t need this,” and that manufacturing a need is the wrong goal; treating the principled objections as real commitments rather than messaging targets; and recognizing that trust, once it’s been spent across a whole industry, isn’t rebuilt by the newest entrant insisting it’s different. It’s worth noting, too, that not everyone in this group is even a committed non-user — about one in ten volunteered that they use chatbots now and then, and last week just happened to be an exception, a nuance the blanket label “non-user” hides completely.

With non-users, listening honestly means hearing past the concerns available in a checklist to the feeling underneath it: a frustration with the world of AI that is genuine, largely undirected, and not going to be talked away. Understanding it — where it comes from, who it’s aimed at, and what would actually earn back a little trust — is the work. The natural next step is to study it properly, in more depth and in more languages, rather than to explain it away.

This concludes the three-part series. Part one introduced the everyday chatbot user; part two mapped five distinct segments.