ChatGPT, Google Gemini, and DeepSeek now absorb nearly 90 percent of all the time people spend inside AI assistant apps, according to Sensor Tower’s State of AI 2026 report covering the first quarter of 2026. That concentration is tightening even as the broader category explodes: global time spent on generative AI apps hit 17.2 billion hours in the first half of 2025 and is projected to reach 36 billion hours in the corresponding period this year. For the dozens of smaller AI assistants still competing for attention, the math is brutal.
Why the 90 percent lock-in changes the AI assistant market
The raw number tells a story about habit formation. When three products capture nine out of every ten minutes users spend with AI assistants, switching costs start to compound. Users build conversation histories, customize settings, and train themselves on specific interfaces. Each session deepens the friction of leaving. That dynamic gives ChatGPT, Gemini, and DeepSeek a structural advantage that goes beyond brand recognition or feature parity.
The practical effect is a widening gap between the top tier and everyone else. Smaller AI apps are not just losing market share in a static pool. They are losing share while the pool itself is growing fast. Sensor Tower projects that global time spent on generative AI apps will more than double year-over-year. That means the absolute hours flowing to non-top-three apps could still grow, but the relative share keeps shrinking. Developers building the next AI assistant face a question that echoes the early smartphone era: can a new entrant break through when users have already settled into daily routines with a small number of dominant products?
A reasonable expectation is that this concentration will accelerate paid-tier conversion inside the leading three apps. Users who spend the most time with a product are the likeliest to hit free-tier limits and pay for upgrades. If 90 percent of engagement sits inside three apps, those three will capture a disproportionate share of subscription revenue, too. The flip side is rising uninstall rates for competitors that fail to earn repeat sessions. The next quarterly data release should show whether this two-tier engagement curve is already taking shape.
For enterprise buyers, the lock-in effect may be even stronger. Once a company standardizes workflows, compliance reviews, and internal training around a specific assistant, the perceived risk of switching grows. That encourages vendors to bundle more services-coding help, document analysis, meeting notes-into a single assistant, tightening the loop between time spent and dependence on a chosen platform.
Sensor Tower’s Q1 2026 data and what it actually measures
The concentration figure comes from Sensor Tower’s State of AI 2026 report, distributed through PR Newswire and published on the firm’s own channels. The report tracks time spent across AI assistant apps, with ChatGPT, Google Gemini, and DeepSeek together accounting for nearly 90 percent of that total in Q1 2026. Separately, the same report notes that ChatGPT has reached 1 billion monthly active users, a threshold that places it among the most widely used consumer apps in any category.
The usage trajectory adds context to the concentration story. Time spent on generative AI apps totaled 17.2 billion hours in the first half of 2025. Sensor Tower projects that figure will climb to 36 billion hours in the first half of 2026, a year-over-year increase of more than 100 percent. That growth rate reflects both new users entering the category and existing users spending longer sessions inside their preferred apps. The three dominant products are absorbing most of that expanded engagement.
Each of the top three occupies a distinct position. ChatGPT, built by OpenAI, has the largest user base and the longest track record in the consumer market. Google Gemini benefits from deep integration with Android devices and Google’s search infrastructure. DeepSeek, which gained rapid traction after launching its open-weight models, has drawn heavy usage particularly in markets outside the United States. Their combined dominance is not the result of a single advantage but of three different distribution strategies converging on the same outcome: persistent daily engagement that crowds out alternatives.
The distribution of the report itself underscores how closely the AI boom is tied to established information pipes. Sensor Tower’s materials circulate through logged-in portals such as newsroom dashboards and regional publishing hubs like the EMEA portal, ensuring that the same headline numbers reach investors, marketers, and product teams at the same time. That synchronization helps lock in a shared narrative about who is winning the AI assistant race.
Gaps in the data and what to watch next
The Sensor Tower report, while specific about the 90 percent figure, leaves several questions open. The State of AI 2026 landing page does not publish the full methodology for how “AI assistant apps” are defined or whether the time-spent metric covers mobile apps only or includes web-based usage. That distinction matters. ChatGPT and Gemini both see heavy browser-based traffic that may not appear in app-level tracking. If the 90 percent figure reflects mobile sessions alone, the actual concentration across all platforms could be either higher or lower depending on how web usage distributes.
Country-level breakdowns are also absent from the publicly available materials. DeepSeek’s usage patterns skew heavily toward certain regions, and without geographic detail, it is difficult to know whether the 90 percent figure holds uniformly or reflects an average pulled by one or two high-volume markets. Session frequency and user overlap data, which would clarify whether the same people are using multiple top-three apps or sticking to one, have not been released in the public summary.
Another blind spot is task-level segmentation. “Time spent” bundles together wildly different behaviors: a student drafting essays, a developer debugging code, a customer asking about a bank statement. If the top three assistants dominate some use cases but not others, niche competitors may still find viable footholds by focusing on specialized workflows, compliance guarantees, or integrations that the big platforms have not prioritized.
Regulatory scrutiny could also change the trajectory. Competition authorities in major markets are already studying how app store policies, default settings, and bundling practices shape user behavior in mobile ecosystems. If AI assistants become default features of operating systems, browsers, or productivity suites, regulators may eventually ask whether those defaults unfairly entrench incumbents. Any requirement to unbundle assistants or offer clearer choice screens could loosen the grip of the current top three.
On the product side, the next signal to watch is multimodal usage. As assistants move more deeply into voice, video, and real-time collaboration, the definition of “time spent” may need to expand beyond text chat sessions. An assistant that quietly summarizes meetings or drafts code in the background might deliver significant value without generating long visible sessions, potentially undercounting its impact in app-based metrics.
For now, though, Sensor Tower’s numbers point to a market that is consolidating faster than many expected. The combination of explosive overall growth and highly concentrated engagement means that new AI assistants will have to do more than match features. They will need a distribution edge-through hardware, enterprise deals, or deeply embedded workflows-to pry minutes away from ChatGPT, Gemini, and DeepSeek. Until that happens, the 90 percent lock-in is likely to remain the defining fact of the AI assistant era.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
