AI Image Models Outperform Chatbots in App Downloads

Releases of advanced image-generation models have become the primary engine for mobile application growth, significantly outperforming traditional conversational AI updates. According to fresh insights from app intelligence firm Appfigures, these visual-centric features generate 6.5 times more downloads than standard chatbot model enhancements.

This trend represents a pivot in user behavior. Previously, demand was largely dictated by text-based conversational upgrades or the addition of features like voice chat interfaces. Today, the ability to generate imagery is proving to be a much stronger catalyst for new user acquisition.

Performance Metrics Across Major Platforms

The impact of these releases is visible in the download spikes for major players like Google and OpenAI:

• Google Gemini: The introduction of the Gemini 2.5 Flash image model last August triggered over 22 million additional downloads in the following 28 days, a jump of more than 4x compared to previous periods.
• ChatGPT: Following the launch of its GPT-4o image model in March of last year, the app saw an incremental increase of over 12 million installs within 28 days. This performance was roughly 4.5x higher than the download volume seen for GPT-4o, GPT-4.5, and GPT-5 releases.
• Meta AI: The release of the “Vibes” AI video feed in September 2025 resulted in an estimated 2.6 million incremental downloads in its first month, demonstrating that visual content—whether static or video—is a powerful draw for users.

Download Spikes vs. Revenue Growth

While interest in new visual AI features is high, Appfigures notes that a surge in downloads does not automatically correlate with increased consumer spending. Many users appear to install these apps out of curiosity to test new capabilities, rather than an intent to commit to paid subscriptions.

For example, the Nano Banana release for Google Gemini drove significant download volume, yet generated only $181,000 in estimated gross consumer spending during its 28-day window. Similarly, Meta AI’s Vibes feature failed to produce meaningful revenue despite its impact on installs.

In contrast, OpenAI has successfully translated this attention into financial performance. The company’s 4o image-generation model launch was linked to an estimated $70 million in gross consumer spending within the first 28 days compared to its established baseline.

The report also examined DeepSeek, though it noted that the platform’s performance represents an outlier. The January 2025 release of DeepSeek R1 resulted in 28 million downloads, marking a significant breakout moment for the company. Unlike the visual model releases, this surge was driven by industry curiosity regarding the efficient training techniques used by DeepSeek, rather than new image-generation capabilities.