Summary of Key Findings
A recent survey on AI usage among American adults reveals significant differences in user income distribution across various AI products: Claude is predominantly used by high-income individuals, Meta AI by lower-income groups, while products like ChatGPT and Gemini have a user base closer to the national average. These disparities are not due to users consciously choosing a particular “class label” but are determined by the pricing strategies and customer acquisition models of the AI services. This is merely an early stage phenomenon in the widespread adoption of AI; as competition among manufacturers intensifies and the market evolves, user distribution is likely to become more balanced.
I. Claude and Meta AI: Two Extreme User Profiles
The survey data highlights these stark differences: 80% of Claude users come from families with annual incomes of over $100,000 (compared to only 50% of all American adults), whereas 32% of Meta AI users come from families with annual incomes of less than $50,000 (only 7% of Claude users fall into this category). It might seem like “rich people use Claude, and poor people use Meta AI,” but this is not a matter of class preference; rather, the products themselves filter out certain user groups.
II. It’s Not You Who Choose the AI; the AI’s Business Model Chooses You
The reason for these extreme differences lies in the “acquisition cost” of the products:
- Claude’s Barriers: There is no truly free version (the free tier offers limited usage, and advanced features like coding and design require payment); the Pro version costs $20 per month, with the Max plan costing between $100 and $200 per month. Users must also register on the official website and use a credit card, which excludes those who are unwilling to spend, do not actively search for such services, or do not have a credit card.
- Meta AI’s Advantage: Meta AI is integrated into social apps like Facebook and Instagram that users use daily, eliminating the need for additional downloads or registrations. This “zero-barrier” approach naturally attracts a wider range of lower-income users who are less likely to seek out new tools.
III. The Middle Ground: Most People Use AI in a More Ordinary Way
Don’t be misled by these extremes; the majority of people use AI in a more mundane way. Products like ChatGPT, Gemini, and Copilot have user incomes that are similar to the national average: 56%-64% of their users come from high-income families (compared to the baseline of 50%), and 15%-22% from low-income families (compared to 24%). For example, ChatGPT was downloaded 770 million times globally last year, indicating that its users are typical Americans with regular internet access, not a specific group. The use of AI is not strongly associated with social class.
IV. The Cycle Between Ability and Income: Will Better AI Be Exclusive to the Rich?
There is a concern that more advanced AI tools might become exclusive to the wealthy. Claude is expensive and performs well in tasks like coding and complex writing. If high-income users can use these tools to increase their efficiency and earn more money, could this create a cycle where “the richer you are, the better you can use AI, and the better you can use AI, the richer you become”? However, for now, there’s no need to worry too much; this is just an early stage.
V. This Is Just a Temporary Snapshot; Things Will Change in the Future
This survey reflects the initial phase of AI adoption:
- Manufacturers Adjusting: Companies like Anthropic are lowering prices and making free features available; OpenAI has added image generation to its free version of ChatGPT; Google has integrated Gemini into its search results. All are working to reach a wider audience.
- Regional Differences: The situation in the United States may not apply in China, where mainstream AI tools are generally free, and user differentiation is more about whether to use AI or which specific tool to use.
In summary, the current user distribution represents a transitional state rather than a fixed pattern. As competition intensifies, the barriers to using AI will decrease, and class differences may gradually fade away.
Finally, the choice of AI tool is not just about personal preference but also reflects the product design and acquisition costs. However, this is temporary. The ultimate goal of AI should be to make high-quality tools accessible to everyone.