A new survey finds that 27% of U.S. adult internet users have social and emotional interactions with artificial intelligence systems, with a third of those users considering their AI bot as a friend.
I wouldn’t blink at either one of those statistics in isolation.
But…both black and Latino groups, particularly black, typically have lower levels of educational attainment in the US.
I would expect that to mean that there is either some really high concentration in one area—like, a exceptionally high level of black users with graduate degrees are using AI, enough to throw the numbers that far from overall proportions in society, or you have different strong “clusters” of different groups using AI—like, there is a really high level of black AI users and a really strong cluster of high-education users.
considers
It does say that users are disproportionately below 50, and that would explain some of it. Younger generations are both more educated and less white than older ones.
Would need to go dig up demographics for each cohort to figure out what proportional numbers would be, but just glancing at that, that’s definitely not going to explain nearly all of that. The groups are roughly equally-sized for younger age cohorts, so that’s be a fair bit of that on age, but there’s still gonna be a substantial education split.
For the 2021-2022 school year, this has whites being granted about six times as many masters or doctorates as either black or Latino students, so that’ll be a pretty big educational gap even for younger age cohorts. Yeah. Huh.
drums fingers
Well, I dunno. It’s an interesting pair of figures, though.
I think you’re on to something. I think the next logical question to ask is - “What proportion of post-graduate non-Whites land in what kinds of roles?” Is there a reason this demographic is using AI companions more that correlates to the kinds of jobs they’re doing?
That was one thing that I wondered about for a moment, like, maybe you have people working very specific types of fields—I’d seen some papers in, like, psychology and stuff talking about human-AI interaction, and maybe this is a really hot thing to be doing to be doing in that field—but this survey was specific to “companion” uses, which I don’t think that would likely include professional use the way it was defined here:
The people who have used LLMs for social, emotional and personal purposes are known as “AI companion users” and a representative sample of 1,000 of them were surveyed to discover how they use their social bots, the nature of their relationships and the impact of their encounters with these systems.
I don’t think it’s professional use either. I’m suspecting structural racism - the people drawn to companion AIs are likely highly educated but unable to use their education fully in their current role. Just my gut speculation. But I’m curious if there’s enough data to show that kind of correlation.
These are some wild numbers, looking at them together.
So, basically, it’s heavily disproportionately loaded towards high-education users. 25% users have post-graduate degrees.
https://www.ccdaily.com/2025/09/datapoints-educational-attainment-in-the-u-s/
This has 14.9% of the US age 25 and older falling into that group.
But it’s also disproportionately loaded towards black and Latino users. This has 30% of the userbase being non-Latino black.
https://blackdemographics.com/black-population-in-the-u-s-reaches-51-6-million/
That’s 12.7% of the US.
I wouldn’t blink at either one of those statistics in isolation.
But…both black and Latino groups, particularly black, typically have lower levels of educational attainment in the US.
I would expect that to mean that there is either some really high concentration in one area—like, a exceptionally high level of black users with graduate degrees are using AI, enough to throw the numbers that far from overall proportions in society, or you have different strong “clusters” of different groups using AI—like, there is a really high level of black AI users and a really strong cluster of high-education users.
considers
It does say that users are disproportionately below 50, and that would explain some of it. Younger generations are both more educated and less white than older ones.
searches
https://www.pewresearch.org/short-reads/2019/07/30/most-common-age-among-us-racial-ethnic-groups/
Would need to go dig up demographics for each cohort to figure out what proportional numbers would be, but just glancing at that, that’s definitely not going to explain nearly all of that. The groups are roughly equally-sized for younger age cohorts, so that’s be a fair bit of that on age, but there’s still gonna be a substantial education split.
searches
https://nces.ed.gov/FastFacts/display.asp?id=72
For the 2021-2022 school year, this has whites being granted about six times as many masters or doctorates as either black or Latino students, so that’ll be a pretty big educational gap even for younger age cohorts. Yeah. Huh.
drums fingers
Well, I dunno. It’s an interesting pair of figures, though.
I think you’re on to something. I think the next logical question to ask is - “What proportion of post-graduate non-Whites land in what kinds of roles?” Is there a reason this demographic is using AI companions more that correlates to the kinds of jobs they’re doing?
That was one thing that I wondered about for a moment, like, maybe you have people working very specific types of fields—I’d seen some papers in, like, psychology and stuff talking about human-AI interaction, and maybe this is a really hot thing to be doing to be doing in that field—but this survey was specific to “companion” uses, which I don’t think that would likely include professional use the way it was defined here:
I don’t think it’s professional use either. I’m suspecting structural racism - the people drawn to companion AIs are likely highly educated but unable to use their education fully in their current role. Just my gut speculation. But I’m curious if there’s enough data to show that kind of correlation.