1. Headline & intro
AI image models have quietly become the strongest growth engine in the mobile app stores, outpacing even big chatbot upgrades. But there’s a catch: downloads are exploding while revenue mostly is not. That gap says more about the future of consumer AI than any benchmark chart.
In this piece we’ll look at why visual models are suddenly driving 6× more installs than text-only upgrades, why only OpenAI is really turning that into cash, and what this shift means for app makers, incumbents like Google and Meta, and users in Europe and beyond.
2. The news in brief
According to TechCrunch, citing new Appfigures data, releases of AI image (and visual) models are now the main driver of growth for consumer AI apps. Appfigures estimates that launches of image models generate about 6.5 times more downloads than traditional chatbot/model updates.
Google’s Gemini app reportedly saw over 22 million additional installs in the 28 days after launching its "Nano Banana" image model in August 2025, boosting downloads by more than fourfold in that period. OpenAI’s ChatGPT app added more than 12 million incremental installs in the month following its GPT‑4o image model launch in March 2025, significantly outperforming earlier text‑focused model releases.
Meta’s AI assistant also saw a bump of around 2.6 million downloads after rolling out its Vibes AI video feed in September 2025.
However, only ChatGPT translated this surge in attention into meaningful in‑app spending: Appfigures estimates roughly $70 million in additional gross consumer spending in the month after GPT‑4o image launched. Gemini’s Nano Banana drove roughly $181,000 in extra spending, while Meta’s Vibes saw no notable revenue impact.
3. Why this matters
The headline is simple: users don’t download AI apps for better reasoning; they download them to make things they can see and share.
The fact that image and video models drive 6.5× more installs than pure chatbot upgrades tells us two things:
Utility is boring, spectacle converts.
Incremental gains in accuracy or latency are invaluable for power users, but invisible to the average person. A new image model, by contrast, instantly yields meme‑ready content and social bragging rights. That’s acquisition rocket fuel.The AI camera is beating the AI notepad.
The winning consumer interface for AI looks less like a smart search box and more like an upgraded camera and editing studio. This shifts where product teams invest: prompt engineering and memory are important, but the real growth levers are visual UX, filters, and sharing workflows.
Who benefits?
- OpenAI clearly knows how to price and package visual AI. Turning a 28‑day image‑model promo into an estimated $70M is evidence of product‑market fit with paying users, not just tourists.
- Large platforms like Google and Meta gain cheap distribution. Even if current revenue is low, they’re seeding accounts and data for future monetization (cloud, ads, bundles).
Who loses?
- Smaller chatbot‑only apps that can’t afford GPU‑hungry image models are getting drowned in the noise. If you’re just “ChatGPT, but with a different color,” you have no story when users now expect eye‑catching visuals.
- Investors chasing MAU vanity metrics will misread these spikes. Downloads without retention or ARPU are 2014‑era app‑store mistakes, now repeating in AI.
The core problem: the features that drive the most virality—unlimited image generation, playful video—are also the most expensive to run and the hardest to put behind a paywall without killing growth.
4. The bigger picture
This trend doesn’t come out of nowhere; it’s part of a broader shift to multimodal AI as entertainment infrastructure.
Over the last two years we’ve seen:
- OpenAI’s GPT‑4o and Google’s later Gemini releases collapsing text, voice, and vision into a single model.
- Meta shipping generative image and video tools inside Instagram, WhatsApp and Facebook.
- A wave of “AI avatar,” “AI girlfriend,” and AI video editing apps temporarily topping app‑store charts whenever they add a new visual trick.
We’ve been here before. Think back to FaceApp, Prisma or Snapchat’s gender‑swap and baby filters. Each new visual effect produced a spike in installs and social feeds full of transformed faces. Today’s AI image models are that phenomenon industrialised: instead of a single filter, you get an infinite creative toolbox.
The difference now is cost and complexity. A viral AR filter is cheap to run; a state‑of‑the‑art diffusion or video model is a GPU furnace. That’s why Appfigures’ numbers matter: they show visual AI can flood the funnel but doesn’t yet translate into sustainable revenue—unless you have OpenAI‑level pricing power, enterprise upsell, or both.
Competitively, this entrenches the biggest players:
- Google can treat consumer image models as marketing for its cloud AI stack.
- Meta can monetise indirectly through engagement and ads rather than direct subscriptions.
- OpenAI can push users from free image fun into paid tiers, API usage and enterprise deals.
Independent app developers are being nudged into one of two roles: niche specialists (e.g. tools for designers, marketers, or specific verticals) or thin UI layers on top of someone else’s models. The days when you could build a unicorn around a single “magic photo” feature look numbered.
The industry signal is clear: AI is no longer only about answering questions; it is becoming the underlying engine of user‑generated content.
5. The European / regional angle
For European users and companies, the rise of image‑driven AI apps collides directly with regulation.
The EU AI Act and the Digital Services Act (DSA) push platforms to label AI‑generated content, include provenance signals and manage risks around deepfakes and disinformation. If your growth comes from people creating faces, political imagery or synthetic news clips, you now carry heavier compliance overhead in the EU than in many other markets.
That cuts both ways:
- For US incumbents, Europe becomes a higher‑friction, lower‑margin region for visual AI roll‑outs.
- For European startups, there is an opportunity to bake compliance, watermarking and rights‑management into their products from day one and sell that as a differentiator.
We already see European‑linked players in this space: Stability AI’s image tech, photo‑editing successes like Lensa and PhotoRoom, design platforms in Berlin and Paris building AI‑first creative workflows. The next generation of these apps will have to integrate consent management, copyright handling and safe‑by‑default presets to satisfy GDPR and upcoming AI‑specific rules.
Culturally, European users are more privacy‑sensitive and more cautious about putting their faces into black‑box systems. That may dampen the kind of “upload all your photos for fun transformations” behaviour that fuelled early US‑centric virality. Expect slower, but more sustainable, adoption curves—and more interest in on‑device or EU‑hosted models.
6. Looking ahead
Over the next 12–24 months, expect three things to happen.
The AI camera moves into the OS.
Apple, Samsung and others are already moving AI enhancement and generative features closer to the system camera and gallery. As this matures, standalone “AI image” apps will struggle unless they offer deep professional or community features, not just effects.Pricing experiments will intensify.
The Appfigures data screams “monetization gap.” To close it, we’ll see:- Limited free image quotas with paid boosts.
- Creator‑tier subscriptions with higher‑res, faster rendering, or commercial rights.
- Enterprise bundles where visual AI is just one feature baked into productivity suites.
Most experiments will fail; a few will define the default business models for the next five years.
Regulation will reshape product design.
As the EU and other regions refine rules around synthetic media, apps will need clearer disclosures, content provenance APIs, and tooling for rights holders. Apps that ignore this will find themselves geo‑blocked, down‑ranked or buried in legal overhead.
Unanswered questions remain: Will users keep paying for image generation once novelty wears off? Can startups afford to compete with hyperscalers on GPU cost curves? And will we see a backlash as social feeds fill with AI‑generated photos and videos that erode trust?
For now, the safe bet is that visual AI will become an expected default capability—like a camera filter—while the business value shifts toward workflow integration, brand tools and enterprise content pipelines.
7. The bottom line
Image and video models are now the biggest growth hack in consumer AI—but, for most apps, they’re still a terrible business. OpenAI is the exception that proves the rule: it converts spectacle into subscriptions. Everyone else is subsidising GPU‑heavy fun for users who rarely pay.
If you’re building in this space, the real question isn’t “How do I add image generation?” but “What do users do with those images that’s valuable enough to charge for?” Until someone answers that at scale, we’ll see more download spikes than sustainable AI businesses.
