Subscriptions Pile Up, But There's No Framework
ChatGPT, Claude, Gemini — the subscriptions keep piling up, but many people find themselves staring at the billing screen each month with no real framework for deciding which tool should handle which task. It's even more true for solo operators. Without a team, your tools are your workforce, and when you subscribe to whatever everyone else says is good, you end up with more subscriptions and less conviction.
A new market study on AI assistant usage offers a useful reference point. Drawing on a large sample, it measures which platforms people actually choose, how they divide work between them, what they trust, and what they're willing to pay for.
About the Study
The paper, by Jennifer Zou, surveyed a representative sample of 1,999 adult AI assistant users in the United States in June 2026. The researchers set out to answer four questions: which platforms users choose, how they split tasks across platforms, how they judge providers' trustworthiness, and how much they value data-handling features. The estimates were weighted against external adoption benchmarks to match the broader population of AI users, so this reads less like the opinion of one segment and more like a map of the market as a whole.
What the Study Found
First, the market is concentrated overall but fragmented underneath. ChatGPT is the primary assistant for 58% of users and Gemini claims 25%, yet smaller platforms hold onto defensible niches. Claude, with just 7% overall share, still captures a third of all coding tasks. Task allocation turned out to be organized by platform rather than by individual user, and technical use dropped off sharply with age.
Second, trust comes from hands-on experience, not reputation. In head-to-head comparisons among people who had used both platforms, Claude was rated the most trustworthy every time — and it also had the widest gap between how users and non-users rated it, the largest of any platform in the study.
Third, privacy concern is nearly universal, but what actually drives behavior isn't how worried people are — it's how much they know. In a choice experiment, the single item users were willing to pay the most for wasn't a better model at all, but a feature that keeps humans from ever reading their conversations, worth $11.20 a month. The more sensitive the task, the more they were willing to pay.
What This Means in Practice
The implication for solo operators is clear. The instinct to standardize on a single tool actually runs against how the market really behaves. People already split their work across platforms by task. The approach that matches the data is to first break your own core work into pieces — coding, writing, research — and then assign the right tool to each one.
The finding on trust is just as practical. If trust is built through use rather than word of mouth, there's little reason to subscribe based on buzz alone. It makes more sense, and matches what the study found, to run candidate tools against the same real task for a set period and judge from there. And for anyone handling customer data, the finding that users place the highest value on blocking human access to their data is worth applying directly to how you design privacy into your own service.
A Few Caveats
This is a survey-based study of adult users in the United States as of June 2026. Market share and usage habits in Korea may differ, and given how fast this market moves, the numbers should be read as a snapshot in time rather than a fixed picture. Even so, the three underlying patterns — task-based division of labor, trust built through use, and privacy behavior driven by knowledge rather than worry — are a solid enough starting point for building your own criteria for choosing tools.
