What Your AI Conversations Are Really Worth

OpenAI announced ads are coming to ChatGPT. I have been waiting for this moment. Not anxiously awaiting, more like inexorably waiting. I started my career out of college in the world of Amazon Advertising and Pay-Per-Click (PPC) machine learning optimization. Seeing the profound impact that AI can have on targeting, optimization, and conversion, I knew it was just a matter of time until this came to AI. In this article, I want to break down a few things:

Why OpenAI is better positioned than most to succeed in advertising Why OpenAI is making this move now, and what it can mean for their overarching business model What it means for the AI market as a whole (the future of shopping)

1. Why OpenAI is better positioned than even Amazon to succeed in advertising

Amazon revolutionized advertising through data, tracking clicks, browsing patterns, and purchase history to predict what you’ll buy next. The retail giant became famous for using predictive analytics so sophisticated that it could identify pregnant customers through their shopping patterns, as detailed in a 2012 New York Times investigation into Target’s similar predictive models. But imagine if they had access to something even more powerful: a text field where customers voluntarily shared their deepest fears, aspirations, and daily struggles. That’s ChatGPT.

ChatGPT announced on January 16, 2026, that they will begin testing ads on free and “Go” subscribers (their $8/month tier). I have been waiting to become one of their lucky test subjects this week, probably because I am #TeamClaude and cancelled my premium ChatGPT subscription a few months ago. From the announcement, ads look fairly predictable and distinguishable. The LLM responds to your query, allegedly without being influenced by paid media, then draws a clear (but thin) line, slaps the magical word “sponsored,” and gives you a product or service recommendation related to your search.

You got skirt steak in the fridge? Why not make some carne asada, and by the way, here’s some fire hot sauce you can buy.

As of now, it is fairly immediate for a user to distinguish between organic LLM generation versus a paid generation. Don’t you see that UI-friendly line? But will it always be the case? How far away are we from companies being able to bid on the weight of a specific set of tokens (casually, the set of tokens that spell out Coca and Cola in the same sentence about refreshing beverages)? This would be similar to PPC advertising, but now you are no longer playing with positioning on digital real estate but rather the conversations attributed to our intelligent artificial companions.

This is besides the point. The format of ads will change, they changed on Google, Amazon, and all platforms alike. Why does OpenAI have a major advantage compared to the Amazons of the world? When you look at the most common use cases of AI in personal life (not enterprise use cases), a few themes emerge: companionship, life planning, life purpose. We tell, talk, and show ChatGPT very intimate elements about our lives. When my clients ask me how to leverage AI to create new and personalized customer experiences, I always bring up two things: intelligence and memory. Intelligence is knowledge of a customer, and memory is simply the ability to store incremental signals over time to strengthen that intelligence.

I recently went into the memory log of my ChatGPT account. The things it remembered about me? My recent health diagnosis through analyzing my heart echocardiogram (my doctor was being slow giving me an interpretation of the results), my detailed financial standing (ChatGPT doesn’t take 3% of my returns like most advisors), and even knew of my problematic relative (I needed some witty responses prepared ahead of the traditional debates at Christmas dinner). This information is not noise, it’s signal. It goes beyond where I clicked, how long I spent looking at that product, or how I got to Amazon in the first place. ChatGPT knows who I am, it knows my passions, my fears, my habits, intelligence about me as a customer, and it has a great memory of it.

These two factors alone position ChatGPT to be able to reinvent digital advertising as we know it.

2. Why OpenAI is making this move now, and what it can mean for their overarching business model

Why now? Let’s go back to September 2024 when OpenAI first announced “Shopping”, the Agentic Commerce Protocol with Stripe that enabled instant checkout within the AI chatbot. By late November 2024, they rolled out shopping research features, and by April 2025, shopping capabilities had expanded globally. It’s impossible for me not to see these shopping announcements and the January 2026 ads rollout as inherently linked. We serve you ads, and you can buy from here, bingo.

But why? What happened to Sam Altman when he said in May 2024 that advertising was a “last resort” and that “ads plus AI is sort of uniquely unsettling to me”? The answer is more nuanced than simple survival, it’s about scale and optionality.

OpenAI’s business model is incredibly tied to tokens, but the challenge isn’t necessarily that subscriptions are unprofitable, it’s that they only capture a fraction of the value being created. Tokens are what LLMs intake as input and what they generate as outputs. Grammatically, a token is fairly close to a syllable. Practically, it’s the unit of measurement, the currency for intelligence. Altman even pointed out that users who are consistently saying “please” and “thank you” to ChatGPT are costing OpenAI millions. Because tokens are expensive!

The math: You may pay $20/month as a ChatGPT Plus user, but how much do you actually use it? The latest GPT-4o model costs OpenAI roughly $2.50 per million input tokens and $10 per million output tokens. For heavy users consuming hundreds of thousands of tokens monthly through extended conversations, complex tasks, or AI agent workflows, that $20 subscription becomes tight on margins. The more you use it, the more context you give it, the more jobs you get AI to complete, the closer you edge toward break-even territory.

Yes, token prices have decreased dramatically, GPT-4o costs about 50% less than GPT-4 Turbo did just a year ago. But the demand for tokens from users has increased exponentially. We are no longer just chatting with AI; now we have teams of AI agents completing work for us, consuming vastly more tokens.

Now here’s the real issue: OpenAI has committed over $1.4 trillion to AI infrastructure over the next eight years and is burning through an estimated $8 billion annually. But let’s be clear, this cash burn isn’t just about serving existing users. It’s R&D, it’s training new models, it’s building infrastructure for future scale, it’s talent acquisition in the most competitive hiring market in tech. This spending exists whether ChatGPT has 100 million or 800 million users. The unit economics of subscriptions might actually be fine, even good. The problem is different.

OpenAI has 800 million weekly active users, but only 5% are paying subscribers. That means 760 million people are using ChatGPT, deriving value from it, leaving traces of their preferences and behaviors, and OpenAI is capturing exactly $0 from them. Subscriptions work, but they’re leaving massive value on the table.

Add to this the pressure of preparing for what many expect to be one of the largest tech IPOs in history. Investors don’t want a company with a single revenue stream, no matter how promising. They want diversification, they want proof of multiple monetization paths, they want to see that OpenAI can extract value from its entire user base, not just the 5% willing to pay upfront. Token-based subscriptions can coexist with advertising and commerce. In fact, for a company of OpenAI’s ambitions, they must.

The economics of AI aren’t forcing OpenAI to abandon their current business model, they’re forcing them to expand it: advertising and commerce aren’t replacing subscriptions, they’re monetizing the 95% who won’t subscribe.

3. What it means for the AI market as a whole (the future of shopping)

OpenAI releases shopping, then it begins testing ads. OpenAI is moving into the consumer world, not just of AI, but of products and experiences. I don’t believe we are that far away from AI agents completing our buying for us. I am emphasizing buying instead of shopping here because I do think AI agents will likely serve a specific purpose for shoppers in the early game, a buying purpose, for utilities, for things that are repetitive and inconsequential for most: toilet paper, hangers, fertilizer sticks for your monstera, a replacement portafilter for your espresso machine. Not shopping, not treating yourself to a fancy watch for your promotion or the Hermes bag of your dreams (will an agent have its own purchase history at Hermes and Rolex?).

Buying is the largest part of the consumer market by volume. AI companies have raised billions and are now pressured to deliver trillions in revenue. Token economics, highly linked to adoption, is either not fast enough or just NOT ENOUGH. It’s time to enter the economics of conversion and digital real estate, but in a new way. Now powered by both customer intelligence and memory, AI companies have the opportunity to reinvent buying and advertising.

The race is on. Will Google with their Gemini and Amazon with their stake in Anthropic be able to leverage their decades of experience in advertising to maintain their edge? Amazon is already rolling out “Buy For Me” agentic shopping. Perplexity launched their shopping assistant in late 2024. The entire AI landscape is converging on the same realization: advertising and commerce are not just ancillary revenue streams, they are essential to survival.

What we’re witnessing is the birth of a new advertising paradigm. Not the banner ads of the 2000s, not the targeted social media ads of the 2010s, but something fundamentally different: conversational commerce powered by unprecedented customer intelligence. The platforms that win won’t just know what we click, they’ll know why we hesitate, what we value, and what we’ll need before we know ourselves.

The question isn’t whether AI will transform advertising. That ship has sailed. The question is who will own this transformation, and at what cost to our privacy, our autonomy, and our ability to distinguish between genuine assistance and sophisticated persuasion. OpenAI just fired the starting gun. Every other AI company is scrambling to their marks.

---

Disclaimer: The perspectives shared in this article are entirely my own and do not reflect the views of my employer, clients, or any organizations I’m affiliated with. References to specific companies and technologies are used for illustrative and analytical purposes only. This content should not be interpreted as investment advice, financial guidance, endorsements, or predictions of who will win the AI race. Also, despite ChatGPT apparently knowing everything about me, it did not write this article, I did.

References

Hill, Kashmir. “How Target Figured Out A Teen Girl Was Pregnant Before Her Father Did.” Forbes, February 16, 2012. OpenAI. “ChatGPT Ads Testing Announcement.” January 16, 2026. OpenAI. “Shopping Feature Launch.” September 2024. Altman, Sam. Statement on advertising as “last resort.” May 2024. TechCrunch. “Sam Altman says OpenAI has $20B ARR and about $1.4 trillion in data center commitments.” November 6, 2025. Fortune. “OpenAI says it plans to report stunning annual losses.” November 13, 2025. HSBC Research. “The Risks Facing OpenAI and its $1.4T in Spending Commitments.” November 2025. Perplexity. “Shopping Assistant Launch.” Late 2024. Amazon. “Buy For Me Agentic Shopping Announcement.” 2025.