by Cosmin Ene
In platforms like ChatGPT, only 3% of users are paying customers- leaving the remaining 97% generating usage and cost, without revenue.
Generative AI (GenAI) is advancing faster than any platform shift in recent history. Billions are being invested, thousands of startups have launched, and AI agents are rapidly becoming embedded in how we write, code, design, and research. But one critical piece remains unresolved: how humans and agents interacting with one another will make money.
Unlike traditional software, where a product can be built once and sold to millions at near-zero marginal cost, GenAI outputs are produced on-demand, uniquely, and expensively. Each answer, image, or conversation consumes computational resources and incurs real-time costs. Monetizing that usage, especially at scale, is proving difficult for most participants in the GenAI ecosystem
Worse still, the conventional models aren’t mapping cleanly. Advertising doesn’t align well with how people use GenAI. Subscriptions work for a few power users, but most people won’t pay monthly for occasional or narrow use. And while venture capital continues to pour in, investor patience is finite. If monetization fails to catch up, the industry risks entering a new kind of AI winter- driven not by lack of innovation, but lack of revenue for the many.
This chapter explores the monetization paradox in business-to-consumer GenAI, where millions use the tools, but few pay. It argues that we’re in a structural transition period, much like the early internet before paywalls, or the music industry before iTunes. With economic pressures mounting, business models must evolve to reflect how people want to engage today: low friction, choice, and clear value.
If they don’t, GenAI’s most promising tools could fail before they have a chance to mature - and in doing so, alter the ethical and economic shape of the digital world to come.
For decades, online services have been monetized through two primary models: advertising and subscriptions. But GenAI’s structure and usage patterns break both.
● Advertising ≠ Alignment
Advertising works best when users browse, scroll, or search, creating space for recommendations and interruptions. But GenAI is different. It’s goal-oriented. You ask, it answers. You move on.
In the traditional search business, you have enough surface to show multiple ads, but in AI, this is different. The promise of AI is precision - you ask a question and get an answer. The world is moving from “Search” to “Find,” and there is not enough of - or not as attractive - a surface to show ads. This precision undermines the ad model. When AI gives users exactly what they want on the first try, there’s no “click path” left to monetize. And while platforms like Snapchat’s My AI, Bing Chat, and Google’s Search Generative Experience are experimenting with sponsored suggestions, this approach comes with serious risk. If users begin to feel their AI is “serving the brand” instead of serving them, trust erodes, and trust is core to AI’s value proposition.
Even in the best case, GenAI simply doesn’t generate enough engagement volume to support traditional CPM-based models. And even if it did, advertising margins are notoriously thin, often fractions of a cent per impression. That math doesn’t work when each query costs multiple cents to process.
● Subscriptions = Commitment Fatigue
Subscriptions are the other go-to model in digital services. But they rely on regular, predictable engagement - something that doesn’t map well to how people use AI.
The GenAI landscape is fractioned. According to ThereIsAnAIforThat, as of 22 August 2025, the current count stood at 40,248 services and applications. There is an AI for everything, from writing CVs, generating art, composing music, tutoring math, helping with legal forms, or even livestreaming endless AI-generated SpongeBob episodes, turning any Wikipedia page into a fake podcast, and yes… generating deepfake memes of a very pregnant Travis Kelce. Most users rely on these tools episodically, jumping between services as needed. Committing to a monthly plan for each one quickly becomes unmanageable. And with the rise of specialized agents, the Web, apps, and software will eventually become truly atomized. While to the user, the frontend may remain intact, in the background, we will have agents who are orchestrated to provide their individual, specialized service to the user.
We’ve already seen the limits of subscriptions play out in real time. According to the reports, only 3% of ChatGPT users pay for the service. The rest generate compute costs - often significant - without contributing revenue. This model may work for a small set of power users, but it cannot sustain an entire industry.
And the macroeconomic picture compounds this issue. As the economy tightens, users hesitate to subscribe, businesses scrutinize every invoice, and households cut back on recurring costs - even for services they like. In this context, asking users to “marry” a tool they only need occasionally feels like overreach.
Ads will work for some, Subscription models will work for a few, but both models together will likely not be enough.
When new technologies break existing economic models, there’s often a limbo phase - what you might call a monetization transition period. We’re in one now.
Think back to the music industry. Napster proved that people wanted digital music, but there was no legal, monetizable way to meet that demand. It took Apple’s iTunes and the 99¢ song to close the gap - aligning usage, pricing, and trust. A similar shift happened with journalism: ads alone couldn’t fund online news, and while the sector is still struggling, a mix of paywalls, memberships, and subscriptions emerged in an attempt to replace lost revenue.
GenAI is facing its version of this gap. Usage is skyrocketing, but monetization is flat. This creates massive operational pressure, not just for startups, but for every platform and infrastructure provider supporting them. I’m convinced that we’ll see a pattern like in other industries such as news media.
Think about yourself. Are you seeing yourself subscribing to dozens or hundreds of AI products and services? You are likely to subscribe to one or two main services like ChatGPT and Perplexity and you will want to occasionally use individual applications which serve a specific need or an in-the-moment interest. For example to book travel, helping your kids understand math, or to have some fun, here’s what’s 18-24 months out:
To move forward, we need new models. A few are already taking shape:
These approaches all reflect a larger shift in consumer psychology: people want flexibility. They want choice. They don’t want to commit to a $20/month tool to solve a $1 problem. They’ll pay - but they want to pay on their terms. Which raises a fundamental question: which model delivers on that?
One increasingly promising answer is pay-as-you-go: letting users pay for what they use, when they use it.
Pay-as-you-go, or microtransaction-based monetization, offers a third monetization path - one that more naturally aligns with how people interact with GenAI today. Most AI tools are used sporadically: for a specific task, a short burst of creativity, or a moment of productivity. People don’t want to subscribe to dozens of tools they only use occasionally, nor do they want to be sold to mid-conversation. What they do want is access, on demand, with control over when and how they spend.
This model isn’t new. We’ve seen how small, on-demand payments helped unbundle music from albums, articles from magazines, and rides from car ownership. The logic is the same here: let users pay in small increments - say, 1¢, 10¢ or 50¢ - for individual AI responses, image generations, or content unlocks. Let them accrue value over time, then settle up when it makes sense.
Case Study: DeepakChopra.ai. DeepakChopra.ai is a wellbeing coach - a Generative AI companion built around the teachings and voice of Deepak Chopra, physician, author, and the worldwide leading expert in mindfulness. The platform delivers personalized conversations, guided meditations, reflections, and wellbeing advice - designed to help users reduce stress, sleep better, and explore mindfulness on demand.
To better reflect real usage patterns, the team implemented a flexible monetization strategy. Alongside a traditional subscription, they introduced a pay-as-you-go option: low-cost “time passes” that give users temporary access without requiring a monthly commitment.
In just four months, 55% of all users started engaging with the pay-per-use model provided by Supertab and 19% of them converted to subscriptions. This hybrid approach allowed the platform to monetize both casual and frequent users, without adding friction.
Of course, microtransactions come with their own design challenges: pricing, user education, and minimizing friction. But these are solvable problems. What they offer in return is a monetization model that’s not just technically viable, but ethically coherent - one that can scale across use cases and economies without compromising trust or accessibility.
Monetization isn’t just about profitability - it’s about priorities. The way GenAI is funded will directly shape its behavior, accessibility, and ethical impact.
If only ad-funded or subscription-based platforms survive, the ecosystem will skew toward optimizing attention, not utility. We’ll see dark patterns emerge: AI tools nudging users toward overuse, upsells, or data-sharing. Tools for education, local languages, or low-income communities may never get built, simply because they don’t fit the monetization model.
Responsible AI requires business models that work for everyone: flexible enough for casual users, sustainable for providers, and inclusive and affordable enough to support diverse users around the world. If we fail to build those models now, we risk locking the future of AI into an extractive, inequitable mold.
The Generative AI revolution is real - but fragile. Without monetization models that reflect how people actually want to engage, the industry may stall before it matures (and it gets into a state of oligopoly, like what happened with the Internet revolution.)
And in doing so, it may shape the next generation of digital tools around scarcity, centralization, and exclusion.
We have a narrow window - perhaps 18 to 24 months - to get this right. That means treating monetization not as a secondary concern, but as foundational infrastructure. Like safety, fairness, or privacy, revenue models shape what gets built, who gets served, and what incentives govern the system.
There may be no perfect model, but we know the current models aren’t enough. Ads and subscriptions certainly have their merits, but where their limitations start begins the land of pay-as-you-go. Between ads and subscriptions lies a more adaptable, user-aligned alternative - pay-as-you-go systems that reward value, respect autonomy, and scale sustainably.
The future of AI won’t be shaped by what it can do, but by how it gets paid, and who it ultimately empowers.
© 2026 Cosmin Ene. All rights reserved.
AUTHOR:
Cosmin Ene grew up in Romania, where bananas were exotic, ice cream was rare, and books were censored, then immigrated to Germany, where suddenly there was freedom, knowledge, and choice. That contrast shaped his belief in building technology from the consumer’s perspective. As Founder and CEO of Supertab, he’s rethinking how we buy content and AI services, not only with ads or subscriptions, but in ways that fit real life: simple, fair, just for a few cents, and mostly cheaper than a banana.
Thank you for exploring AI for Humanity, a project built by humans, powered by AI, and guided by values. Join us in shaping a more human‑centered future.
Adobe Firefly, Canva Magic Studio, ChatGPT, Gamma.app, Gemini Notebook, Otter.ai, Perplexity, and Suno.
Copyright © 2026 American Society for AI (ASFAI) and The International Social Impact Institute® (The ISII®). All rights reserved.
The American Society for AI is a non-profit and the preeminent organization for Artificial Intelligence (AI).
Our mission is to create a better world with AI.
Learn how information is handled and review the terms governing use of this platform:
Part 4: Finance, Technology & Investments