AI Search Has a Wealth Gap Most Marketers Haven't Priced In
The GEO conversation has been loud. Every conference, every newsletter (including this one), every SEO consultant has spent the last year telling you to optimize for AI search visibility. Get cited by ChatGPT. Show up in Google's AI Overviews. Make your content AI-friendly. Good advice, mostly.
But there's a demographic assumption baked into all of it that nobody seems to want to talk about: AI search users are disproportionately wealthy.
Reflect Digital's data, published this week on MarTech, lays out the UK numbers. Among households earning £25,000 to £30,000, about 18% use ChatGPT regularly. At £70,000 to £80,000, that number jumps to 49%. At £100,000 and above, it sits between 48% and 58%.
The curve isn't gradual at all. More like a cliff.
The US data tells the same story
The UK numbers aren't an outlier. A Brookings nationwide survey of American adults found that 34% of people earning over $100,000 use AI in their professional lives. For those earning under $30,000, that number drops to 9%. Nearly a 4x gap.
Education tracks the same pattern. 67% of respondents with a bachelor's degree or higher have used generative AI for personal purposes. Among high school graduates, it drops to 46%. For daily usage, the gap widens further: 20% of college-educated users versus 8% of those with a high school diploma. And the age split reinforces it. The 30-to-44 cohort leads professional AI adoption at 31%, while workers over 60 sit at 8%.
This probably isn't surprising if you think about it for a minute. AI tool adoption runs on three things: access (having a device and a subscription), capability (knowing what to do with it), and confidence (trusting the output enough to change your behavior). All three correlate with income and education in predictable ways. FutureDotNow found that 52% of working-age adults in the UK can't complete all essential digital tasks required for work.
These are not people switching to ChatGPT as their default search engine.
Why your GEO investment is a de facto demographic filter
This connects directly to budget decisions. When you optimize content for AI search visibility, you are optimizing for an audience that skews high-income, college-educated, and probably under 45. You're not optimizing for "everyone who searches." You're optimizing for a specific demographic slice.
For luxury brands, financial services, SaaS, and B2B tech, this might be exactly the audience you want. Great. Your GEO investment just got a demographic tailwind you didn't even have to pay for.
For everyone else, the math gets uncomfortable. If you sell to a mass-market consumer audience, if your customer base includes people earning $40,000 to $60,000, your GEO-optimized content is reaching a narrower slice of your addressable market than your traditional SEO efforts. And you might be reallocating resources toward the narrower channel without realizing it.
I keep coming back to this: the AI search traffic that does arrive converts at a higher rate. Superlines' data shows AI-referred visitors convert at 14.2% compared to Google's 2.8%, and spend 68% more time on site. But those numbers are inseparable from the demographics of the people generating them. High-income, high-education users convert better at basically everything, across every channel, for almost every product category. The conversion lift traces to the audience composition, not the search interface. Selection bias in a new outfit, and honestly, a pretty convincing one.
Three audiences, not one
Reflect Digital's research identifies three distinct user segments emerging from the fragmentation. AI-first users who delegate tasks and decisions to chatbots entirely. AI-assisted users who cross-reference AI output with traditional search before acting. And AI-avoidant users who stick with Google, social media, or direct navigation.
The mistake I see most teams making is treating "search" as a single channel with a new sub-channel bolted on. It's more like the audience split into three rooms, each with its own demographics, intent signals, and content tolerance. The AI-first room skews affluent and confident. The AI-avoidant room includes a significant chunk of the population you might actually need to reach.
We've already covered how AI Overviews are eating click-through rates unevenly across categories. Add the income dimension and the picture gets more complicated: the traffic you're losing to AI Overviews was reaching a broad demographic. The traffic you're gaining through AI citations reaches a wealthier, more educated one. The net effect isn't neutral even if the raw numbers look similar.
What I'd actually pull up this week
Open your Google Analytics for the last 90 days. Separate your AI referral traffic (filter for chatgpt.com, perplexity.ai, claude.ai, gemini.google.com referrers) from your standard organic Google traffic. Compare the two audiences on whatever demographic or behavioral proxies you have. Average order value, if you're in ecommerce. Lead quality score, if you're B2B. Geographic distribution works too, because metro vs. rural often correlates with income in ways that show up cleanly in analytics.
If the AI referral audience looks meaningfully different from your organic audience on any of those signals, that's information you need before you shift more resources toward GEO.
The benchmark I'd watch: AI referral traffic currently represents roughly 1% of total website traffic for most sites, growing about 1% per month. At current growth rates, that could reach 5-8% of total organic traffic by year end. If the income skew holds, a meaningful and growing share of your organic reach will effectively be earmarked for people who already convert at the highest rates. At that level, the demographic composition of that traffic becomes a strategic question, not a footnote.
From what I've seen, most teams haven't even set up the tracking to separate these audiences. They're running GEO experiments without knowing who they're actually reaching. That seems backwards.
When "everyone is doing GEO" hides the real tradeoff
93% of AI search sessions end without an external website click. Traditional Google search? That zero-click rate is 34%. So even within the high-income cohort that uses AI search, the overwhelming majority never visit your site. They get the answer inside the chatbot and move on.
The counterargument is that brand mentions in AI responses build awareness even without clicks. And that's probably true, to a degree. But brand awareness among affluent early adopters is a very specific outcome to be optimizing for, and I think it's worth naming it plainly rather than pretending GEO is a universal traffic strategy.
I think GEO is worth doing. I've said that before and I mean it. But the honest framing is that it's a premium audience play disguised as a search strategy. If you know that going in, you can make smart allocation decisions. If you don't, you risk shrinking your addressable market while your analytics dashboard shows improving conversion rates. Everything looks like it's getting better when really it's getting narrower.
The difference between a good GEO strategy and a wasteful one isn't the tactics. It's whether you know who you're actually reaching, and who you've quietly stopped reaching.
