That viral Ghibli trend helped ChatGPT make millions of dollars

OpenAI recently released its Images 2.0 model, and it has already kicked off a new trend that lets you imagine yourself in the same frame as your younger self. And this is not the first trend sparked by new image models, with earlier ones like the Ghibli trend set off by ChatGPT or the 3D figurine trend fueled by Google’s Nano Banana. According to experts, these phenomena also led to the fastest growth in apps, which, in turn, helped generate revenue faster than any other AI model releases.

AI-based image generation and manipulation contribute a large share to the growth of AI chatbots based on large language models. This was recently reported (via TechCrunch) by Appfigures, an app store optimization and analytics platform.

Although new model releases with conversational features, complex problem-solving, and voice-based features previously drove growth for AI chatbots, image-related tools have largely steered downloads for these apps. According to Appfigures’ data, ChatGPT accrued nearly 12 million new downloads in just four weeks after it introduced the viral image generator powered by its GPT-4o model last year in March, which set the internet on fire with its Ghibli image trend. Similarly, Google’s Gemini gained over 22 million downloads after the release of Nano Banana, based on the Gemini 2.5 Flash model, last year in August.

According to the platform’s findings, the introduction of image and video-centric features and models, on average, resulted in a 6.5 times the usual increase in the number of downloads. OpenAI’s Ghibli phenomenon led to 4.5X the growth ChatGPT otherwise witnessed during milestone launches, such as that of GPT-4o, GPT-4.5, GPT-5.

Besides the additional download figures, OpenAI and Google also generated revenue from paid models, Appfigures claims. OpenAI, which saw enough demand to eventually limit image generation to paid models, is believed to have generated around $70 million in revenue over the first 28 days after launching the image model. In contrast, Google, despite accumulating nearly twice as many downloads, is estimated to have made only about $181,000. Notably, Google also offered free AI subscriptions to billions of users, especially in growing markets such as India, which may have impacted its revenue.

One exception was the release of the popular open-source model DeepSeek R1 back in January 2025. It was the only text model with more downloads than ChatGPT and Gemini during their image download rush phases, and gained enough traction for its special features that made it more attractive than ChatGPT and Gemini.