I recently attended and spoke at an AI conference. One of the things that became clear to me is that we are probably headed for a new digital divide related to AI. What do I mean by that?
In that short period of time, AI has touched a large majority of computer users. A survey from Pew in September 2025 showed that 95% of adults had heard of AI. At the time of the survey, 47% of people had heard a lot about AI, up from 26% measured in a 2022 survey. That’s bound to be a higher number in the summer of 2026. As you might expect, people with a high awareness of AI tend to be younger and better educated. For example, 62% of adults under 30 had heard a lot about AI, compared to 32% of those 65 and older. 60% of adults with post-graduate degrees had heard a lot about AI, versus 38% of those with a high school diploma or less.
Only about 3% of U.S. households pay for a consumer AI subscription. There is a lot of expectation that AI companies will be forced to greatly increase the costs of monthly access, which will undoubtedly lower the percentage of folks who are comfortable working directly with AI. Most people who use AI today interface through some tool like virtual assistants, GPS navigation apps, streaming algorithms, or weather forecasts. According to a survey from Quinnipiac University from March 2026, 51% of respondents say they use an AI tool, up from 37% a year earlier.
Another Pew poll showed that 46% of Americans hold a negative view of AI, while only 26% view it favorably. 57% believe the risks outweigh the benefits, and 41% actively distrust AI systems.
All of these statistics point to a possible AI-divide, separating those who believe in and use AI, from those who don’t trust AI and refuse to use it. This divide already exists, and the future question that will have to be answered is if those who distrust AI will eventually be won over. It’s easy to forget that AI has only been available to the public since the end of 2022, which, for a new technology, is still in the infancy stages.
There will be consequences if there is a permanent gap between those who use AI and those who don’t or won’t. If AI brings measurable advantages to those who use it, then, over time, non-users will be at a disadvantage in many ways. There is also the possibility that those who distrust AI will be able to thwart its expansion. There are already numerous communities creating barriers to new data centers.
It’s easy to conclude that using AI is a choice, and that anybody who doesn’t use AI has to live with that decision. But consider a few statistics from the National Center for Education Statistics (NCES). It reports that 21% of adults are functionally illiterate, meaning they cannot complete basic reading tasks, such as filling out job applications, reading medicine instructions, or helping with children’s homework. 54% of adults have reading skills below a sixth-grade reading level. Less than half of adults demonstrate strong reading comprehension skills.
These statistics are relevant when talking about using and benefitting from AI, which is ultimately a language model. Literacy is far less of an issue when talking about the use of broadband, because it doesn’t require advanced reading skills to watch videos and engage in other forms of online entertainment and social media.
It’s way too early in the life of AI to draw any conclusions about a possible divide, but it’s not hard to foresee a likely divide between those who benefit from AI and those who can’t or won’t. In a decade, this might become the most important digital dividing line, more so than the digital skills divide we talk about today.
Doug, I think you’re identifying a very real issue. As someone who would probably be considered one of the “older” users of AI, I find myself relying on it every day, not because it’s a novelty, but because it makes me better at my job.
I use AI to review contracts, improve emails, research complex topics, analyze financials, build dashboards and spreadsheets, and even keep track of important themes throughout the month that it later helps me turn into board presentations. That alone has probably cut my preparation time in half. More importantly, AI remembers connections between conversations, ideas, and documents that I would never be able to keep straight on my own. I don’t see it as replacing my thinking. I see it as augmenting it.
One observation I’d add is that I don’t think the divide will simply be between younger and older generations. In our organization, I have employees nearing retirement who have embraced AI and are using it to stay ahead of the game. At the same time, I have younger employees who barely touch it. The more meaningful divide may be between those who intentionally learn how to use AI well and those who don’t, regardless of age.
I also think the literacy discussion deserves a little nuance. AI may actually reduce some traditional literacy barriers because it can explain, summarize, translate, and simplify complex information. The bigger long-term differentiator may not be reading ability itself, but discernment: knowing how to ask good questions, evaluate AI’s answers, recognize when it’s wrong, and combine its capabilities with sound human judgment.
Broadband connected people to information. AI is beginning to connect people to understanding and productivity. Communities and organizations that learn to leverage it thoughtfully will likely have a significant advantage over those that simply choose to ignore it.
I would say there are multiple divides coming together, and not the one the old farts in congress think. ie, connectivity isn’t in the way here, you can run claud or chatgpt on the lowest cost budget cell plan.
-there’s an idealogical divide, those that will use it, and those that will reject it.
-there’s a financial divide, those that have the means to pay for tokens enough to be productive, and those that never get above free/cheap plans.
-nations that have the means to build, and thus export controls making for a 1st/3rd world type divide
-product divide, the line between the serious operators from point 2, and consumers that aren’t building with AI, just consuming trained output. This would be AI personal assistant users vs the companies that make them.
And then massive risk of collapse as new methods to build ‘hardware’ obsolete the old ones. As the divides I describe above build, they could then collapse almost overnight.
AI is most likely to bring chaos, without ever becoming conscious or malicious, just destroying economic models faster than people can replace them.