← Back to feed

A new Epoch AI/Ipsos survey makes the income stratification of AI adoption unusually legible. Across three pooled waves of 5,000 US adults surveyed between March and April 2026, the gap between Claude and Meta AI is not a gap — it is a chasm.

Eighty percent of US adults who used Claude in the previous week live in households earning $100,000 or more a year. The national figure is 50%. Only 7% of Claude’s weekly users live in households earning under $50,000, against a national average of 24%.

Meta AI sits at the other end of the distribution. Just 37% of its weekly users are in $100K+ households. Thirty-two percent earn under $50,000 — nearly five times the rate of Claude users, and well above the national baseline.

Every other major platform clusters in a relatively narrow band between these two extremes: Microsoft Copilot at 63.7%, ChatGPT at 60.3%, Grok at 56.2%, Google Gemini at 55.9%. Claude is the outlier at the top; Meta AI is the outlier at the bottom. The middle is where everyone else is fighting for the same user.

Two different theories of where AI money comes from

The income split is not accidental. It reflects genuinely different go-to-market bets.

Anthropic has built Claude as a premium product — API-first, enterprise-forward, and priced to signal quality rather than access. Its user base looks like its positioning: high-income, likely professional, likely paying. The company’s revenue model depends on this segment and its enterprise contracts. If 80% of your weekly users earn over $100K, you are disproportionately in the hands of people who expense software.

Meta AI is doing the opposite. It is distributed free through the world’s largest social and messaging platforms — WhatsApp, Instagram, Facebook, Messenger — and does not require a separate account, subscription, or download. That distribution strategy reaches parts of the income distribution that paid AI products do not, and Meta’s numbers reflect it. The strategy is volume and integration, not premium positioning.

What the middle looks like

The cluster of ChatGPT, Gemini, Copilot, and Grok between 56-64% in $100K+ households is worth noting on its own terms. These are not mass-market tools in the Meta AI sense — they still skew well above the national income average. But they are also nowhere near Claude’s concentration.

ChatGPT’s 60.3% figure is interesting given its position as the largest platform by weekly usage in the survey. It has scale and income concentration simultaneously — the biggest absolute number of high-income users of any platform, by virtue of its user base size, even without the per-user skew of Claude.

Grok’s 56.2% figure is notable given that xAI has leaned heavily into a politically conservative positioning that might have suggested a different income profile. The data does not show that.

Caveats that matter

The Claude sample is small. With only 201 pooled respondents, the 90% confidence interval on the $100K+ figure runs from 75.2% to 84.4% — wide, but still dramatically above every other platform. The directional finding is clear even if the precise number is uncertain.

The survey measures self-reported weekly usage, not revenue or enterprise contract data. High-income users are not necessarily paying users, and vice versa. Anthropic’s API business and enterprise contracts bring in substantial revenue from organizations that may not appear in household income surveys at all.

The data covers the US adult population only. Global usage patterns, particularly in markets where Meta’s platforms dominate, would look different.

Source: Epoch AI/Ipsos survey, waves 1-3 (March–April 2026). Data and methodology published at epoch.ai.