About half of American adults never use an AI chatbot. Pew Research Center figures, reported by The Register's Dan Robinson, put non-users at 51 per cent of all US adults, which makes users a slight minority of the population.
The gap is widest among older adults. Roughly three-quarters of those aged 65 and over say they never use chatbots, as do 58 per cent of those aged 50 to 64. About four in ten adults under 50 report not using them at all. Those aged 18 to 29 are the heaviest users at 66 per cent, and about a third of that group still say they never use one.
Non-users were offered five possible reasons. Six in ten cited simple lack of interest as a major factor. Concern about data privacy and about the accuracy of chatbot answers came next. Fewer than a third said they did not know how to use them, and fewer than one in five worried about what other people might think of them for using one.
The forward-looking numbers are firmer still. Of the non-users, 67 per cent said they were unlikely to use a chatbot in the next 12 months, and of those, 40 per cent said they were not at all likely to change their mind. Twelve per cent were unsure. Five per cent thought they would probably try one in the coming year.
Pew offers no figures on trust beyond the accuracy concern, and the survey's authors are not quoted in the report.
The survey therefore measures something the industry rarely treats as a category at all, which is the decision not to participate.
Read as a product manager would read them, the figures describe a remarkably healthy proposition. Market share, 51 per cent. Twelve-month forward retention, 95 per cent, since only one non-user in twenty expects to defect. Of those intending to stay, two in five describe themselves as immovable.
There is no onboarding. There is no infrastructure. There are no inference costs, no capacity planning and no model updates. Abstention has never hallucinated, never gone down and never shipped a regression. It has no accuracy complaints, for the straightforward reason that it does not answer.
The only weakness in the proposition is that nobody has yet established how to bill for it, which is also the single problem the rest of the category does not have.
The Department of Unnecessary Research notes that the leading product in American artificial intelligence has no servers, no roadmap and no revenue, and that 40 per cent of its users have formally described themselves as not at all likely to leave.
It is the only offering in the sector with a confident twelve-month forecast, and the forecast is nothing.