What Generative AI Adoption Statistics Straight From the Horse’s Mouth Say About AEO/GEO and Whether You Need to Be Doing It
Generative AI adoption for 18-24 year-olds and 25-34 year-olds is almost double that of the next largest user group, 35-44 year olds. For older generations, it’s fractional by comparison. That gap is worth knowing if you’re wondering how much of a part of your marketing AEO/GEO should be this year.
This data, which came straight from OpenAI/ChatGPT, means that the first, most important question for marketers in 2026 is does my target audience use AI in the first place? That can help you prioritize your efforts and assess whether it’s business as usual or whether you need to go nuclear on GEO.
Conveniently, OpenAI just released this batch of generative AI adoption statistics via ChatGPT. It covers use by purpose/type and demographics, with a level of nuance I don’t actually think was entirely intentional.
Adoption By Age
First, the big beats: the data prove what we all already know, which is that adoption of AI has been highest among younger generations. There is a lot of flux in those younger demographics, but static numbers for as recently as March 2026 look like the following:
Share of Total Messages
- 18-24: 30.3%
- 25-34: 35.8%
- 35-44: 18.6%
- 45-54: 9.1%
- 55-64: 3.7%
- 65+: 2.4%
Older user numbers are quite low, overall, but they are on an impressive upward adoption curve. That user base is vastly smaller, but it’s multiplying.
What This Means If You’re Not Marketing to 20-Somethings
All that is to say that if you sell real estate in 55+ communities, you’re probably ok to stop freaking out about AI search visibility right now. But today’s millennial is tomorrow’s boomer, so as we always said with SEO—AEO is like planting trees, the best time to do it is yesterday.
So for anyone in the business of marketing to, well, everyone, the goal should be to establish their business as what’s known in AI-speak as a “stable entity.”
Stable entity: Something with enough data about it available to LLMs to be regarded first and foremost as a “thing,” and eventually as an authentic and trustworthy thing. Can be a company, person, place, concept.
The opposite of that is to remain a string of words, which without additional context, are just that. And to LLMs, those words are math. For a chatbot to calculate the probability of words being in close proximity to your brand name, it needs a lot of data to establish a relational and referential basis.
How LLMs Learn What They Know About You
So how do LLMs “know” what they know about your company? As I explained in my last post on the basics of how AI works [https://sonardigital.co/how-to-optimize-for-aeo-ai-search/], there’s a threshold for how and where that information comes from. I think that dividing line sits roughly at the level of Fortune 500 companies. Above that level, there is so much information about your company available across the internet that your entity may have already been included in training model data.
Below that line, which is pretty much all of us, odds are LLMs have not been trained on your data. So in order to provide any kind of response to a user, they have to rely on retrieval.
But if you’re looking to improve your AEO/GEO, you have to understand how retrieval works before looking into what is actually retrieved. We’ll cover that in our next piece in this series.
For internal teams navigating how AEO/GEO fits into their broader digital strategy, learn more about how we work alongside marketing teams to provide additional strategy and expertise.