Sam Altman once called advertising a “last resort.” On February 9, 2026, that last resort arrived. OpenAI began showing ads inside ChatGPT conversations for users on the free tier and the $8/month Go plan, with ad personalization turned on by default. If you’ve been chatting with ChatGPT about recipes, expect meal kit promotions. If you’ve been asking about career changes, expect job platform ads. Your conversations are now the targeting signal.
The timing made this story impossible to ignore. Just five days earlier, Anthropic had aired a $14 million Super Bowl campaign built entirely around mocking the idea of ads in AI conversations. The tagline: “Ads are coming to AI. But not to Claude.” One day after the Super Bowl, OpenAI proved them right.
But the real story isn’t the corporate theater. It’s what happens to your data when the company running the world’s most popular AI chatbot decides that your conversations are an advertising product.
How ChatGPT Ads Actually Work
OpenAI’s help center documentation describes a system that mines your conversations for advertising signals. Here’s what’s happening:
Targeting data: Ads are matched based on your current conversation topics, your past chats, your prior ad interactions, your general location, and your language settings. If you’re researching recipes, you’ll see grocery delivery ads. If you’re discussing travel, expect hotel promotions.
Personalization defaults: Ad personalization is enabled by default. You have to actively find the setting and turn it off. If you do turn it off, ads don’t disappear - they just become less targeted, falling back to the current conversation thread rather than your full chat history.
Who sees them: Logged-in adult users on the free tier and the $8/month Go plan. Users on Plus ($20/month), Pro, Business, Enterprise, and Education plans don’t see ads.
Where they appear: Ads show up as labeled sponsored links beneath ChatGPT’s responses. They’re excluded from temporary chats, logged-out sessions, image generation responses, and conversations about health, mental health, or politics.
What advertisers get: Aggregated reporting - total views and clicks. OpenAI says advertisers don’t see individual conversations, chat history, names, emails, precise locations, or IP addresses.
The catch: Users who opt out of ads entirely may face stricter daily message limits on the free tier. In other words, refusing to be an advertising product may cost you access to the product itself.
The Privacy Problem No One Is Addressing
OpenAI’s privacy assurances follow a familiar template: we don’t sell your data, advertisers can’t see your conversations, everything is aggregated. These are the same promises that Google, Facebook, and every ad-supported platform made before their data practices became front-page scandals.
The problem isn’t what OpenAI shares with advertisers. It’s what OpenAI itself now has an incentive to collect and retain.
When advertising becomes a revenue source, every conversation becomes a data point. The platform’s economic incentive shifts from “help the user solve their problem as quickly as possible” to “keep the user engaged long enough to show them ads, and learn enough about them to make those ads worth paying for.” This isn’t speculation about OpenAI’s intentions. It’s how advertising economics work everywhere they’ve been deployed.
ChatGPT conversations are categorically different from web searches or social media posts. People tell their AI chatbot things they wouldn’t type into Google. Medical symptoms. Relationship problems. Financial anxieties. Career insecurities. Mental health struggles. When Stanford researcher warned that sensitive information shared with ChatGPT “may be collected and used for training,” they were describing a platform where the data is inherently more intimate than anything Facebook ever harvested from a Like button.
As one privacy analysis noted, unlike social media companies that must infer user data from browsing behavior, “ChatGPT users directly disclose sensitive information - medical conditions, relationship problems, financial situations, career concerns, and mental health struggles.” The inference gap that gave users at least some plausible deniability on traditional platforms doesn’t exist in a conversational interface. You’re telling the ad platform exactly what you’re thinking.
OpenAI says ads won’t appear near sensitive topics like health or politics. But the targeting system still processes those conversations to determine what is and isn’t sensitive. The model has to read your health question to know not to show you a health ad. That distinction - between using data for targeting and using data to exclude targeting - is thinner than it sounds.
The Financial Pressure Behind the Decision
OpenAI didn’t introduce ads because it wanted to. It introduced ads because the math demanded it.
Internal documents obtained by The Information project OpenAI will lose $14 billion in 2026 - roughly three times its estimated 2025 losses. The company plans to spend $200 billion through the end of the decade, with 60% to 80% going to compute costs for training and running AI models. OpenAI’s own forecasts show the company won’t reach positive cash flow until 2030.
With 800 million weekly active users, the vast majority on the free tier, the advertising math is obvious. Sam Altman predicted ChatGPT ads could generate billions in revenue by 2026 and exceed $25 billion by 2030. Those aren’t “last resort” numbers. Those are “core business strategy” numbers.
This is the economic reality behind every promise about privacy protections and labeled ads. When a company is losing $14 billion a year and sitting on hundreds of millions of free users, the gravitational pull toward more aggressive monetization is enormous. Today it’s labeled sponsored links at the bottom of responses. What does it look like in two years when the losses haven’t stopped and investors want returns?
OpenAI already has plans to develop new ad formats, including follow-up questions about sponsored products directly within chat. That’s not an ad at the bottom of a page. That’s the AI itself steering a conversation toward a product.
The Super Bowl War and What It Revealed
Anthropic’s “A Time and a Place” campaign, created by agency Mother and directed by Jeff Low, ran four spots depicting real categories of questions people ask AI assistants - health, relationships, fitness, work - interrupted by absurd sponsored responses. In one ad, a man asks his chatbot “Can I get a six-pack quickly?” and gets pitched fictional insoles. The tagline: “There’s a time and a place for ads. Your AI isn’t it.”
Anthropic CCO Sasha De Marigny framed the choice: “Technology can be a bicycle for the mind…or another surface competing for attention. We want Claude to be the former.”
Sam Altman fired back on X, calling the ads “funny” but “clearly dishonest,” insisting OpenAI would “obviously never run ads in the way Anthropic depicts them.” He accused Anthropic of “doublespeak” for using “a deceptive ad to critique theoretical deceptive ads that aren’t real.” Then came the populist pitch: “Anthropic serves an expensive product to rich people…we also feel strongly that we need to bring AI to billions of people who can’t pay for subscriptions.”
The framing was deliberate. Altman cast Anthropic as an elitist company charging for a luxury product, and OpenAI as the democratic alternative bringing AI to the masses - funded, of course, by mining those masses’ conversations for ad targeting data. It’s the same argument Facebook made a decade ago: the service is free because you’re not the customer, you’re the product.
Anthropic published a blog post stating that conversations with AI assistants are “meaningfully different” from interactions with search engines or social media, noting that their analysis of Claude conversations showed “an appreciable portion involve topics that are sensitive or deeply personal.” The company committed to no sponsored links, no advertiser-influenced responses, and no third-party product placements. Their business model: enterprise contracts and paid subscriptions.
Whether Anthropic can sustain that position is an open question. Adweek noted the parallel to Netflix, which famously promised “no ads” before introducing an ad-supported tier. Anthropic acknowledged this, stating that if they need to revisit the approach, they would “be transparent about the reasons for doing so.” That’s not quite the same as “never.”
The Industry Pattern
OpenAI isn’t alone. Google has reportedly briefed advertisers on plans to bring ads to Gemini in 2026, though Google’s VP of Global Ads Dan Taylor publicly denied it after the report. Perplexity has already experimented with sponsored results in its AI search product.
The trajectory is clear: AI chatbots are following the same monetization path as search engines, social media, and streaming services before them. Free access builds a massive user base. The user base becomes an advertising product. Privacy protections get introduced, then gradually loosened as the revenue pressure mounts. The pattern has repeated so many times it barely qualifies as a prediction.
What makes AI chatbots different - and potentially worse - is the depth of data involved. Google knows what you searched for. Facebook knows what you liked. ChatGPT knows what you were thinking about, worrying about, and trying to figure out, in your own unfiltered words. That’s a qualitative leap in the intimacy of the data available for advertising, and it’s now being monetized by a company that’s losing $14 billion a year and needs to find a way to stop.
What You Can Do
If you’re a ChatGPT free or Go user who doesn’t want your conversations feeding ad targeting:
- Turn off ad personalization. Go to Settings and disable the personalization toggle. This limits targeting to your current conversation only, rather than your full chat history.
- Clear your ad data. OpenAI provides the option to view and delete the data used for ad targeting.
- Use temporary chats. Ads don’t appear in temporary chat mode.
- Consider paying. The $20/month Plus tier remains ad-free. Whether that price stays stable once ads become a major revenue source is anyone’s guess.
- Consider alternatives. Anthropic’s Claude, at least for now, has committed to no ads. Local models running through Ollama or similar tools never touch anyone else’s servers.
The Bottom Line
OpenAI introduced ads into the most intimate digital medium ever created - a conversational interface where hundreds of millions of people share their real thoughts, fears, and questions. Ad personalization is on by default. Opting out may cost you message limits. And the company losing $14 billion a year has every financial incentive to expand this program, not constrain it. If you’re using ChatGPT for free, you’re no longer just the user. You’re the inventory.