Visual AI model launches drive huge download spikes, but revenue lags
Visual AI launches are sending millions to app stores, but most are not turning curiosity into cash. ChatGPT and Gemini won download spikes; only OpenAI’s image push delivered a major spending lift.
Visual AI features are turning model launches into app-store events, but the money trail is far thinner than the download spikes suggest. Appfigures found that image and video model updates drove 6.5 times more incremental downloads than traditional model releases, a reminder that eye-catching features can pull users in fast even when they do not immediately convert them into paying customers.
The clearest examples came from ChatGPT and Gemini. OpenAI added 4o image generation to GPT-4o on March 25, 2025, and Appfigures estimated that ChatGPT picked up 12 million incremental downloads in the following 28 days. Google followed with Gemini 2.5 Flash Image, also known as Nano Banana, in August 2025, and Appfigures estimated 22 million incremental downloads in the same kind of launch window. Meta AI’s Vibes feature, introduced on September 25, 2025 as a feed for discovering, creating and sharing AI videos, brought in an estimated 2.6 million incremental downloads.
But the spending picture tells a different story. Appfigures said ChatGPT’s 4o image generation release added about $70 million in estimated gross consumer spending over 28 days, the largest lift in the analysis. By contrast, Nano Banana generated only $181,000 in estimated gross consumer spending in its 28-day window, while Meta AI’s Vibes launch produced no meaningful revenue. The gap underscores a growing divide in the consumer AI market: features built to spark attention can spread quickly, but only a few are proving they can support durable revenue.
DeepSeek R1 was the biggest spike in Appfigures’ analysis, with an estimated 28 million incremental downloads in the 28 days after its January 2025 release. Appfigures treated that surge as an outlier because it was tied to a global awareness event, not simply a product feature. Even so, the comparison shows how much of the current AI download race is still driven by novelty, headlines and social sharing rather than clear monetization.
OpenAI later said image generation in ChatGPT became one of its most popular features, with more than 130 million users creating more than 700 million images in the first week. Sam Altman has framed image generation as part of building image tools that are “beautiful, but useful,” while Google described Gemini 2.5 Flash Image as a tool for “high-velocity visual creation” and conversational image editing. The market response suggests those features can deliver huge reach, but the harder test for consumer AI remains whether that attention can be turned into sustained revenue.
Know something we missed? Have a correction or additional information?
Submit a Tip
