Image AI models now drive app growth, beating chatbot upgrades

Recent findings from Appfigures reveal that the release of image AI models is significantly boosting growth for mobile applications, achieving 6.5 times more downloads than traditional model updates. This represents a notable evolution from previous trends where updates to conversational models, such as chat interfaces, were the primary drivers of app demand. For example, the introduction of image models by ChatGPT and Gemini has led to substantial spikes in downloads. Specifically, Gemini’s image model, dubbed Nano Banana, resulted in over 22 million new downloads within just 28 days of its launch last August, marking a fourfold increase in app downloads during that timeframe. Similarly, ChatGPT saw an increase of more than 12 million installs following its GPT-4o image model release in March of the previous year, which was approximately 4.5 times greater than the downloads generated by its earlier model updates. Other image model launches have also contributed to increased downloads, albeit to a lesser extent. Meta AI’s video feed model, Vibes, gained around 2.6 million additional downloads after its debut in September 2025. While it’s classified as a video model, its focus on visual content aligns with the growing trend. However, the report underscores a critical point: a rise in downloads does not always equate to higher revenue for mobile apps. While these image model releases entice users to try out the enhanced image generation features, they do not guarantee conversion into paying subscribers. For instance, despite Nano Banana's impressive download numbers, it generated only an estimated $181,000 in gross consumer spending in the month following its release. In contrast, OpenAI’s GPT-4o image model managed to convert interest into revenue, raking in an estimated $70 million in consumer spending during its launch month. Interestingly, Appfigures also examined DeepSeek’s performance, which saw 28 million downloads after its January 2025 release. However, this was not a typical model comparison event; DeepSeek gained notoriety for its innovative training techniques, capturing the tech community's attention in a way that was distinct from image model interest. This case illustrates how curiosity can significantly influence download trends, even when not directly related to image AI capabilities.