Image AI models now drive app growth, beating chatbot upgrades

Image model releases are now the primary growth engine for AI mobile apps, generating 6.5x more downloads than traditional text-based model updates, according to app intelligence firm Appfigures. The shift marks a clear departure from earlier days when chatbot upgrades and voice features drove demand. For example, ChatGPT’s GPT-4o image model added 12+ million incremental installs in the 28 days after its March 2025 launch—roughly 4.5x the downloads seen from its GPT-4o, GPT-4.5, and GPT-5 model releases combined. Similarly, Google’s Gemini Nano Banana (part of Gemini 2.5 Flash) spurred 22+ million downloads in the same window, a 4x lift over its baseline. Even Meta AI’s video-focused Vibes feature added 2.6 million downloads after its September 2025 release, underscoring that visual content consistently outperforms text.

However, Appfigures cautions that download spikes don't always translate to revenue. While ChatGPT’s 4o image model generated an estimated $70 million in gross consumer spending over 28 days, Google’s Nano Banana only produced $181,000 during the same period, and Meta’s Vibes led to no meaningful revenue. The report also highlights DeepSeek’s R1 model as an outlier—28 million downloads came from curiosity about its low-cost training methods, not image capabilities. The takeaway: image models are powerful acquisition tools, but monetization remains uneven, with only established ecosystems like ChatGPT converting users into paying subscribers effectively.