OpenAI launched advertising in ChatGPT this week. The financial reality behind the pivot tells you everything about where AI is heading.

In January 2026, OpenAI announced that ChatGPT would begin displaying ads to free-tier users in the United States. This week, those ads went live, with Target and its retail media division Roundel among the first partners to test contextual advertising inside AI-generated answers. Roundel will test contextual product placement alongside conversational responses, marking the first major retail media network to integrate commerce directly into an AI interface.

The irony is hard to miss. ChatGPT was positioned as the antidote to Google's ad-cluttered search results. A clean interface where you ask a question and get an answer, no sponsored links, no banner ads, no algorithmic manipulation. Anthropic leaned into the contrast during the Super Bowl, running ads that mocked the idea of AI assistants serving you advertisements mid-conversation.

And yet here we are. The company that was going to kill the ad-supported internet just joined it.

The implementation is more considered than the headlines suggest. Ads appear at the bottom of ChatGPT answers when there is a relevant sponsored product or service based on the current conversation. ChatGPT selects which ad to show by matching advertiser submissions against the conversation topic, past chats, and previous interactions with ads.

According to Target's fact sheet, ads are served based on keywords in the user's prompt. A user asking about countertop cooking appliances might see an ad for an air fryer from a pilot participant, clearly labelled and visually separate from ChatGPT's answer. Traffic from ChatGPT to Target has been growing 40 percent on average each month, which explains why the retailer volunteered to go first. Roundel's participation signals that large retail media networks see conversational AI as a critical distribution channel worth testing immediately.

The guardrails are deliberate. Free-tier and ChatGPT Go users see ads. Plus ($20/month), Pro ($200/month), Business, and Enterprise subscribers remain ad-free. No ads appear near sensitive topics including health, mental health, or politics. Users can turn off personalisation and clear the data used for ad targeting at any time. OpenAI has stated explicitly that ads do not influence the answers ChatGPT gives.

Whether users believe that last point will determine how quickly this scales.

The financial picture explains why this was inevitable. OpenAI generated $12.7 billion in revenue in 2025, a staggering number for a company that barely existed five years ago. But it spent roughly $22 billion in the same period, posting an estimated net loss of $9 billion.

The unit economics tell the story more clearly. ChatGPT has 800 million weekly active users, but only 5 percent of them pay for the service. That is roughly 20 million subscribers out of a user base larger than the population of Europe. Compute and technical talent consume approximately 75 percent of total revenue. HSBC analysts project that OpenAI still will not be profitable by 2030 even with advertising revenue included.

Internal documents obtained by The Information show OpenAI forecasting $1 billion in revenue from "free user monetisation" in 2026, growing to nearly $25 billion by 2029. The math is clear: with a 5 percent conversion rate and infrastructure costs that grow faster than subscriptions, advertising is not optional. It is structural.

OpenAI is not arriving at advertising alone. The entire AI industry is converging on the same conclusion.

Google has confirmed it will bring ads to Gemini in 2026, extending its advertising infrastructure into AI-generated responses. Perplexity is already running contextual ads alongside its search answers, generating roughly $20,000 in advertising revenue in 2024 but with ambitions to scale significantly. Amazon is building an AI content marketplace that would position AWS as an intermediary between publishers and AI model builders, another form of paid content access that functions like advertising infrastructure.

The "advertising-free AI" era lasted approximately 18 months. From OpenAI's public launch of ChatGPT in November 2022 to its first ad announcement in January 2026, the window between "we will never show ads" and "here are the ads" was remarkably short by technology industry standards.

This convergence is not surprising when you look at the economics. Inference costs money. Every query a free user sends costs OpenAI significant compute resources, with analyst estimates suggesting costs ranging from fractional cents to several cents per query depending on the model and response length. At 800 million weekly users, the majority of whom pay nothing, the maths does not work without a third-party revenue source. Google figured this out with search 25 years ago. The AI industry just took three years to reach the same conclusion.

This is where the story connects to commerce and payments in a way that most coverage has missed.

Target is not just an early ad partner. Target is also one of the launch retailers for Google's Universal Commerce Protocol and has built a ChatGPT app that lets users browse and shop directly inside the AI. Roundel, its retail media division, manages advertising across multiple surfaces. Target is treating ChatGPT the same way it treats any other media channel: as a place where discovery leads to transaction.

This is the convergence of advertising and payments. The ad is not just awareness. It is the beginning of a transaction that an AI agent can complete. A user asks ChatGPT about cooking appliances, sees a Target ad for an air fryer, and in the near future, buys it without leaving the conversation, with Stripe or PayPal processing the payment through the same agentic commerce infrastructure that Google, Shopify, and OpenAI are building. Both payment networks understand that this represents a fundamental shift in where commerce originates: no longer on websites or mobile apps, but inside conversational interfaces where intent and discovery converge seamlessly.

For payments companies, this matters because it creates a new surface where transactions originate. Every ad inside an AI conversation that leads to a purchase needs a payment rail. The ad networks and payment processors that can bridge the gap between conversational discovery and instant checkout will capture a new layer of commerce that did not exist 12 months ago.

The stakes are substantial. If even ten percent of Target's ChatGPT traffic converts to purchases, the Roundel division could see hundreds of millions in incremental retail media spend flow to OpenAI's ad network within 18 months. Stripe and PayPal will need to ensure their APIs can handle the specific patterns of agentic commerce, where a conversation, a transaction, and a fulfillment request all happen in rapid succession.

The rollout timeline tells us where this is heading. OpenAI is expected to expand beyond the U.S. in Q4 2026, introduce additional ad formats in 2027, and build a self-service advertising platform for smaller businesses by 2027 or 2028. The $25 billion revenue target by 2029 requires scale that only a self-service platform can deliver.

The two-tier question also looms. ChatGPT now has a clear free tier (ad-supported, limited features) and paid tiers (ad-free, full capabilities). This mirrors the model that streaming services, social platforms, and news publishers have already adopted. Whether this creates a meaningful quality gap between the AI experience of paying and non-paying users will determine how sustainable the model is.

And there is the measurement problem. Advertisers can track clicks and impressions on Google. They understand cost-per-click and return on ad spend. Conversational advertising is different. How do you measure the ROI of an ad that appears at the bottom of an AI answer? What is a "conversion" in a conversation? OpenAI will need to build an attribution model that advertisers trust, or the $25 billion target remains a projection on a spreadsheet.

This measurement challenge will determine which ad platforms succeed at scale. Retailers are accustomed to sophisticated attribution from their digital channels. OpenAI's attribution model will need to meet that standard, or we will see CPM and ROAS expectations diverge significantly from what advertisers achieve on Google and Facebook.

If every AI company eventually needs advertising revenue to survive, did we replace one ad-supported internet with another, or are we building something fundamentally different?