ChatGPT’s momentum is still there, but the shape of that momentum is changing—and the change is showing up in a place that matters more than most people think: app retention.
New figures from Sensor Tower, a mobile market intelligence firm, suggest that ChatGPT downloads and usage growth are cooling while uninstall activity is rising. In April, Sensor Tower reports that ChatGPT uninstalls increased 132% year-over-year. The trend appears to have intensified in May, with uninstalls up 413% year-over-year. Sensor Tower also notes that monthly active user growth has slowed: ChatGPT grew monthly active users by 168% in January, but that figure dropped to 78% by April.
Those numbers don’t necessarily mean ChatGPT is “failing.” They do, however, point to a shift from the early phase of explosive adoption—when new users were discovering the product at a rapid pace—to a later phase where competition, switching behavior, and retention become the main drivers of performance. And for OpenAI, which is widely expected to be approaching major corporate milestones including an IPO, the difference between “growing” and “growing efficiently” can be the difference between a story that excites investors and one that raises questions.
What makes uninstall data especially revealing is that it’s not just a measure of interest. Uninstalls are a behavioral signal. They often reflect disappointment, redundancy (“I tried it and moved on”), or substitution (“I found another chatbot that fits my needs better”). In other words, uninstall rates can capture the gap between what users expect from a chatbot and what they experience after the novelty wears off.
The timing is also notable. Sensor Tower’s report links the sharp rise in uninstalls in May to OpenAI’s February deal with the Pentagon. That doesn’t automatically imply causation—there are always multiple factors in play—but it does raise an important question: how do high-visibility partnerships affect consumer sentiment and app behavior? For some users, government and defense involvement can feel like a step toward broader deployment and legitimacy. For others, it can trigger discomfort, privacy concerns, or a sense that the product is shifting away from personal use. Even if only a subset of users react, the effect can show up quickly in uninstall metrics because those users are already installed and already aware of the app.
At the same time, the uninstall spike could also reflect something less political and more practical: the market is now crowded. When ChatGPT first surged, it was one of the few mainstream options that felt genuinely useful. Today, users have more choices than ever—ranging from other AI chatbots to tools embedded directly into operating systems, browsers, and productivity suites. When alternatives are plentiful, users become more selective about what they keep installed. They may download a chatbot for a specific task, test it, and then remove it once the task is done or once they find a better fit elsewhere.
This is where the story becomes more interesting than a simple “growth is slowing” narrative. The early AI chatbot era rewarded acquisition. The next era rewards retention and differentiation. Sensor Tower’s data suggests ChatGPT is still large enough to maintain growth, but the rate at which new users are converting into long-term app retention is weakening.
That weakening shows up in the contrast between January and April monthly active user growth. A 168% increase in January followed by a 78% increase in April indicates that the pool of “easy wins” is shrinking. In growth terms, this is normal: when a product expands rapidly, it eventually reaches segments of the market that are harder to convert. But it also suggests that the competitive environment is changing faster than the product’s ability to sustain the same level of excitement.
There’s another layer to consider: uninstall rates can rise even when overall engagement remains healthy. Some users uninstall because they’re cleaning up apps, switching devices, or reorganizing their home screens. But a year-over-year jump of the magnitude Sensor Tower reports is difficult to explain purely through housekeeping. It implies a meaningful change in user behavior relative to the prior year.
So what might be driving that behavior?
First, expectations have changed. Early adopters often used ChatGPT as a novelty and a curiosity. Later adopters tend to evaluate it more like a utility. If the product doesn’t consistently deliver the kind of results users want—whether that means accuracy, speed, reliability, or usefulness in day-to-day tasks—users may decide it’s not worth keeping installed. They can still access similar capabilities through web interfaces or other apps, so uninstalling becomes a low-cost way to reduce friction.
Second, the “AI assistant” category is fragmenting. Users don’t just want a chatbot; they want an assistant that fits their workflow. Some people want coding help. Others want writing assistance. Others want tutoring. Others want customer support automation. As competitors tailor experiences to these niches, generic chatbots face a tougher challenge: they must either broaden their value proposition or deepen it enough that users feel locked in.
Third, the market is learning how to price and package AI. Subscription models, usage limits, and feature gating can all influence retention. If users perceive that the best experience is behind a paywall, or if they hit limits that interrupt their usage patterns, they may uninstall and return later—or switch to a competitor with a different model. Even if the underlying technology improves, perceived friction can still drive churn.
Fourth, public attention cuts both ways. High-profile partnerships and policy debates can create uncertainty. Even if the product itself doesn’t change, the surrounding narrative can affect user trust. For consumer-facing apps, trust is a retention lever. If users believe the app is becoming less aligned with their values or more entangled with institutions they don’t want to support, they may remove it.
None of this means OpenAI is losing relevance. Sensor Tower’s framing matters: it says ChatGPT still has a substantially larger user base. That suggests the product remains dominant in scale, even if the growth curve is flattening. But dominance at scale doesn’t automatically translate into investor confidence if the market is shifting from “viral adoption” to “sustainable engagement.”
This is where the IPO conversation becomes relevant. Investors typically look for evidence that a company can grow without relying on constant hype. For a consumer app, that means retention metrics, engagement depth, and the ability to convert new users into repeat users. Download and install numbers are useful, but they’re not the whole story. Uninstall rates are a proxy for whether users are sticking around after the initial trial period.
If uninstall rates are rising while monthly active user growth is slowing, it can indicate that the product is still attracting users but not keeping them at the same rate as before. That doesn’t necessarily mean revenue will fall immediately—because revenue can be driven by a smaller subset of highly engaged users—but it can signal risk in the long term. For an IPO, risk perception matters. Markets can tolerate volatility, but they struggle with narratives that suggest the company’s growth engine is weakening.
There’s also a strategic implication for OpenAI: the company may need to treat retention as a first-class product goal, not just a byproduct of model quality. Model improvements matter, but so do product design choices that reduce churn. That includes onboarding that sets correct expectations, personalization that makes the app feel tailored rather than generic, and workflows that encourage repeat use. It also includes transparency around changes—especially changes that affect user experience, such as feature availability, safety policies, or usage limits.
Another unique angle here is that uninstall rates can be influenced by platform dynamics. App stores and mobile operating systems evolve. Users update apps, migrate to new versions, and sometimes uninstall older apps when they no longer see value. But Sensor Tower’s year-over-year comparison helps control for some of that. The fact that the uninstall rate rose sharply suggests something beyond routine platform churn.
It’s also worth noting that the chatbot category is increasingly “always available” through the web. If users can access similar functionality without keeping the app installed, uninstalling becomes less costly. That means uninstall rates might rise even if users remain engaged through other channels. In that scenario, the uninstall metric would overstate churn. However, Sensor Tower’s data is still valuable because it reflects a real shift in how users manage their AI tools. Even if engagement continues elsewhere, the app’s role in the user’s daily life may be diminishing.
So what should we watch next?
The next few months of Sensor Tower data will likely clarify whether the uninstall spike is a temporary reaction or a sustained trend. If uninstalls remain elevated while monthly active user growth continues to slow, it would strengthen the case that retention is deteriorating. If uninstalls normalize and active user growth stabilizes, it could indicate a short-term disruption—perhaps tied to a specific event, update, or external narrative.
Beyond that, it would be useful to see whether ChatGPT’s user base growth is shifting toward different demographics or usage patterns. For example, growth might increasingly come from users who use the service intermittently rather than daily. That can still be profitable, but it changes the risk profile. Intermittent users are more likely to churn when a competitor offers a better experience for a particular task.
There’s also the question of how OpenAI’s product strategy is evolving in response to competition. If competitors are winning by offering more specialized experiences, OpenAI may need to respond with clearer segmentation—either by building distinct modes for different tasks or by integrating deeper into existing productivity ecosystems. The more ChatGPT becomes part of a user’s workflow, the less likely they are to uninstall it.
Finally, the uninstall data invites a broader reflection on what “growth” means in AI. In the early days, growth was about discovery. People downloaded the app because it was new and impressive. Now growth is about habit. The winners in this category will be the ones that turn impressive demos into reliable routines.
ChatGPT has already proven it can capture mainstream attention. The challenge now is to convert that attention into durable usage. Sensor Tower’s figures suggest that conversion is becoming harder. That doesn’t undermine the technology’s