Image generators are now the single biggest growth lever for AI mobile apps, outpacing chatbot upgrades by a wide margin. According to TechCrunch AI, citing fresh data from app intelligence firm Appfigures, image model releases drove 6.5x more downloads than traditional model updates. The shift marks a clear break from the era when voice chat and conversational upgrades drove the install spikes.
What the numbers show
The gap between visual launches and language launches is hard to ignore.
- Google Gemini’s Nano Banana: Added 22+ million downloads in the 28 days after launch, lifting installs more than 4x over the prior baseline.
- ChatGPT’s GPT-4o image model: Pulled in 12 million incremental installs, roughly 4.5x more than GPT-4o, GPT-4.5, and GPT-5 text releases combined.
- Meta AI’s Vibes (video feed): Added 2.6 million downloads in 28 days after its September 2025 launch.
- DeepSeek R1: 28 million downloads after January 2025, though TechCrunch AI notes that was a curiosity-driven breakout, not an image moment.
The pattern is consistent across the major players. People want to see what the new model can draw, not what it can say.
The revenue gap nobody’s talking about
Here’s where the story gets interesting. Downloads aren’t dollars.
Gemini’s Nano Banana drove the largest install spike of the bunch but generated only $181,000 in estimated gross consumer spending over its 28-day window. Meta’s Vibes pulled millions of installs and produced no meaningful revenue. ChatGPT was the lone outlier that converted attention into cash, with its GPT-4o image model driving an estimated $70 million in gross consumer spending versus baseline.
That’s a 387x revenue gap between two products with comparable download bumps. Pricing, brand trust, and existing subscription infrastructure clearly matter more than the wow factor of any single launch.
Why visuals win the install moment
Image outputs are shareable, screenshotable, and instantly understandable. A clever ChatGPT response lives inside the app. A Studio Ghibli portrait or a Nano Banana edit lives on Twitter, Instagram, and group chats. Every shared image is free user acquisition.
Text capability has also hit a comfort zone. Most users feel the major chatbots are good enough at writing. Image quality, on the other hand, still produces visible jumps that anyone can recognize at a glance.
What this means for AI builders and operators
The practical takeaways from the Appfigures data:
- Treat image launches as marketing events, not feature releases. They’re your biggest organic growth channel.
- Don’t confuse top-of-funnel with monetization. A download spike means curiosity, not commitment. Build your conversion funnel before the launch, not after.
- Bundle image generation into existing paid tiers. OpenAI’s $70M result wasn’t magic. It was a mature subscription engine catching the wave. Standalone image apps without that infrastructure left money on the table.
- Watch for the next visual frontier. Video is where the install pull is heading next. Vibes was a small signal. Sora, Veo, and the next wave of consumer video tools will likely repeat this pattern at larger scale.
The broader industry dynamic is clear. The competitive battlefield for consumer AI is moving from text to pixels to motion. Companies that own the social-shareable output own the acquisition flywheel. Companies that own the subscription rails turn that flywheel into revenue.
Full breakdown and charts are available at the original TechCrunch AI report.