The Appfigures report on AI app downloads contains a data point that should concern every AI product team celebrating download spikes from image model launches: additional downloads do not reliably translate into additional revenue, and the gap between download performance and revenue performance across the three major platforms is dramatic. Google's Nano Banana image model drove more than 22 million incremental downloads in its 28-day window, a larger download spike than ChatGPT's GPT-4o image model produced. Yet Nano Banana generated only an estimated $181,000 in gross consumer spending over that same period. Meta AI's Vibes launch added meaningful downloads but no measurable revenue. ChatGPT's GPT-4o image model, by contrast, generated an estimated $70 million in incremental gross consumer spending in the 28 days following its launch compared to its prior baseline.

The $70 million versus $181,000 comparison between ChatGPT and Gemini's image model launches is the most important number in the report, and the gap is large enough that it cannot be explained by differences in download volume alone. The more plausible explanation is that ChatGPT had already built a paying subscriber base and a conversion funnel that transformed curiosity-driven installs into subscription revenue at a rate that Gemini had not yet achieved. Users who downloaded Gemini to try Nano Banana had no established habit of paying for AI services through that app and no clear prompt to upgrade. Users who downloaded or re-engaged with ChatGPT for image generation were arriving in an ecosystem already optimized for converting engagement into paid subscriptions. For AI companies planning image model launches as growth events, the Appfigures data suggests the download spike is the easy part. Converting that spike into revenue requires a subscription infrastructure, pricing model, and user journey that most platforms have not built as effectively as OpenAI.