Platform telemetry can show what a provider's own users do, but it can't show why they chose that platform over another, what they'd pay to change how their data is handled, or where they turn when their default tool won't cut it. Profound’s latest research fills that gap, measuring platform choice, task-level allocation, trust, and privacy valuation across the US AI-using public.
Key findings
- The market is concentrated but not monolithic. ChatGPT (58%) and Gemini (25%) lead primary-platform share, but smaller players hold defensible ground. Claude claims 33% of coding tasks, nearly matching ChatGPT and far ahead of Gemini, despite a 7% overall share, while Copilot doubles its share in work tasks versus personal ones.
- Trust must be earned, not inherited. Claude wins every head-to-head trust comparison against ChatGPT and Gemini, and shows the largest gap between how non-users and actual users rate it, evidence that for challenger platforms, reputation lags well behind the experience of using the product.
- Claude monetizes disproportionately well. With 34% of its users on a paid plan, Claude edges out ChatGPT (32%) and beats Gemini (18%) despite its smaller footprint, pointing to a small but high-intent and deliberate user base.
- Privacy concern is universal; action is not. More than 80% of users worry about how their data is used, but only 18% have ever paid for better privacy protection, and the strongest predictor of protective behavior isn't concern, it's whether users actually know their assistant's training policy.
- Humans, not models, are what users pay to avoid. In a discrete-choice experiment, users valued keeping human reviewers out of their conversations at $11.20/month, nearly 4x what they'd pay to stop training on their data, with willingness-to-pay rising sharply as task sensitivity increases.
Access the full report for the complete platform funnel, task-signature breakdowns, demographic gradients, and the full privacy valuation model.