by Sakshi Dhingra - 5 hours ago - 3 min read
Image-generation upgrades are now doing what chatbot updates used to do: pushing users to download AI apps at scale. New Appfigures data cited by TechCrunch shows that image model releases are generating 6.5x more downloads than traditional AI model updates, marking a clear shift in consumer interest from text-only assistants to visual creation tools.
The biggest signal came from Google Gemini. After the release of its Nano Banana image model, Gemini added 22 million-plus downloads in the following 28 days, lifting downloads by more than 4x during that period. ChatGPT saw a similar pattern after introducing GPT-4o image generation, adding 12 million-plus incremental installs in 28 days.
The reason is simple: users can instantly see the value. A better chatbot model may feel abstract, but an image model gives people a direct reason to install, test, share, and compare results.
The stronger download spike does not automatically mean stronger revenue. Appfigures noted that Gemini’s Nano Banana produced a larger install jump than ChatGPT’s image release, but generated only about $181,000 in estimated gross consumer spending in the 28 days after launch.
That makes the current AI app race more complicated. Image models are excellent acquisition tools, but companies still need to convert viral usage into paid subscriptions.
The wider market data supports the same trend. Sensor Tower reported that users spent more than 48 billion hours in AI-powered apps in 2025, nearly 10x higher than 2023. Downloads reached 3.8 billion in 2025, more than double 2024 levels, while sessions grew 256% to cross one trillion worldwide.
This shows AI apps are no longer being downloaded only for curiosity. Existing users are returning more often, and visual tools are giving new users a stronger reason to join.
ChatGPT remains the dominant AI app globally, but Gemini’s image-led growth shows Google has found a powerful mobile acquisition strategy. Sensor Tower’s 2026 mobile analysis placed ChatGPT first among AI apps, followed by Gemini, DeepSeek, Doubao, and Perplexity.
For Google, the Nano Banana spike matters because it proves Gemini can compete not only through search integration or Android distribution, but through viral creative output.
The next phase of AI app competition may not be decided by which chatbot gives the smartest answer. It may be decided by which app gives users the most shareable, creative, and instantly useful output.
Text models still matter, especially for productivity and research. But for mobile growth, image and video models are proving more powerful because they create visible results, social sharing loops, and faster user curiosity. The real winner will be the company that turns those download spikes into lasting engagement and paid usage.