So, Is the Consumer AI Assistant Market Closed?

The consumer AI assistant market is already concentrated at the level that most users encounter first: general-purpose primary choice. The stronger claim—that the market is closed across tasks, trust, privacy, payment, and the production system behind the interface—is not established by the evidence. The most defensible answer is therefore two-part: the default is powerful and unequal, but important competitive margins remain live and their durability is unresolved.

That is not a compromise designed to avoid a conclusion. It follows from using the right unit for each claim. The survey measures how current US AI users choose platforms, distribute tasks, evaluate providers, protect their data, and trade off privacy features. The BIS study measures a selected set of large AI-producing firms across economies and supply-chain layers. Together they show several forms of concentration, but they do not measure one unified market with one common share.

Concentrated where the default is chosen

The consumer evidence justifies a clear judgment about general-purpose reliance. ChatGPT is the primary assistant for 58.2% of the weighted sample, and Gemini for 25.4%; together they account for 83.6% of reported primary-platform choice. ChatGPT’s lead over Gemini is much larger in primary share than in past-month usage, and adjacent primary rankings are statistically separated down through fifth place. The market’s center of gravity is therefore not an illusion created by awareness alone. Two platforms receive most of the default relationship, while smaller platforms occupy a thin general-purpose middle and lower tier.

This concentration has real consequences. A primary platform receives the majority of a user’s tasks, which can make it the place where history, preferences, files, and expectations accumulate. Routine use becomes easier because the user does not have to reconsider the provider every time. A challenger entering at this layer must overcome the convenience of an existing default as well as whatever product differences users perceive.

But the evidence does not support the phrase “closed” if it means exclusive control over all user activity. Primary choice is a majority category, not a measure of every conversation. A person counted as a ChatGPT primary user may still route coding, work, or a sensitive task elsewhere. About four in five users have tried a second assistant, and about 43% intentionally route different task categories to more than one platform. The second figure is the more demanding evidence of functional competition, but it is still a minority and is based on self-reported allocation across the survey’s six categories.

That combination describes a market with a strong default and a meaningful portfolio of alternatives. The two facts should not be averaged into a vague middle. They answer different questions. The primary share tells us where most routine reliance sits. The multihoming result tells us that a substantial group makes deliberate task-level choices. The first is evidence of concentration; the second is evidence that concentration has not eliminated all competitive discretion.

The strongest overstatement would be to treat the 83.6% figure as 83.6% of revenue, time, profit, or exclusive use. It is the sum of two reported primary shares in a weighted sample of people who already use AI assistants. It says a great deal about the default interface and much less about the value of every task or the future direction of switching. The relevant population is also not all US adults and not the global market. It is the AI-using public represented by the survey’s frame.

The practical question is consequently not whether a challenger can make itself available. It is whether it can win enough repeated, valuable occasions to make the default respond. A platform that is secondary for most users but primary for a difficult task may exert more competitive pressure than its general-purpose share suggests. A platform that is widely tried but rarely used again may exert less. This is why a single market-wide label obscures the mechanism that matters.

Competition survives in tasks, users, and channels

The task evidence prevents the default hierarchy from being mistaken for a complete ranking of platform capability. Claude has 7% overall primary share but 33% of primary coding share, close to ChatGPT’s 39% and above Gemini’s 15%. Copilot’s share roughly doubles in work tasks compared with personal tasks. ChatGPT and Gemini remain generalists, especially in informational use. The contrast shows that a smaller platform can win a recurring job without becoming the user’s primary assistant for everything.

This is more than a statistical curiosity. Task fit changes the cost of substitution. A user who needs a provider for a specialized workflow may be willing to maintain a second platform even when switching all ordinary activity would be inconvenient. If the task recurs, the specialist can accumulate its own familiarity and value. A generalist can still dominate the total relationship while losing a strategically important part of it.

The niche evidence is not a guarantee of durability. The task shares are weighted self-reports from six elicited categories, not a complete account of every workflow. They do not show whether a coding advantage persists after the generalists improve, whether users pay more for it, or whether the niche expands into other tasks. They show that aggregate share and task share are different variables, and that the difference is large enough to matter for how competition is measured.

User composition and payment reinforce the same conclusion from another direction.