OpenAI is now testing ads in ChatGPT for free and low-cost users in the US. Free and Go tier subscribers see ads; Plus, Pro, Business, Enterprise, and Education tiers remain ad-free. This is precisely the moment Anthropic's no-ads pledge was designed to contrast.
The company's published principles sound reasonable enough: ads won't influence ChatGPT's answers, conversations stay private from advertisers, users can opt out of personalization or dismiss individual ads. Infrastructure costs money. Ads help fund broader access. Premium tiers stay clean.
But look at how the matching actually works. OpenAI is serving ads based on your conversation topics, your chat history, and your past ad interactions. Researching recipes? Expect meal kit promotions. When multiple advertisers want the same moment, the company decides which message is "most relevant." This is exactly how every ad-supported product has worked since search engines started selling keywords.
Every Platform Has Made These Promises
Mission alignment: ads fund access. Answer independence: ads don't change responses. Conversation privacy: advertisers don't see your chats. Choice and control: you can manage preferences.
You've heard this before. The question isn't whether OpenAI believes these principles today. It's whether the economics of running frontier models will let them keep believing it in 2028.
"We'll expand responsibly as safeguards mature."
That sentence, buried in the announcement, is the tell. Advertising programs grow. They integrate into revenue targets. They get reexamined when the next earnings call demands more. Guardrails that exist at launch have a way of becoming inconvenient later.
The deeper problem isn't privacy policies. It's incentive structures. An ad-supported product has a fundamentally different relationship with its users than an ad-free one, and those relationships diverge over time.
Right now, ChatGPT's optimization target is helpfulness. Once ad revenue becomes material, there's a second consideration: does this conversation present an opportunity to surface a sponsored message? The objectives sometimes align. When they don't, users can't easily distinguish genuine recommendations from commercially motivated ones. OpenAI says sensitive topics (health, mental health, politics) won't show ads "during our test." That qualifier is doing a lot of work.
OpenAI CEO Sam Altman previously defended the company's ad plans on Twitter, framing ads as necessary to "democratize AI access." The argument sounded strained at the time (March 2025), given OpenAI's $200/month Pro tier and $20/month Plus tier. It sounds even thinner now.
The "democratizing access" framing has a specific function: it positions advertising as a noble mission rather than a revenue decision. But you can't simultaneously claim you're building AI for everyone while charging premium prices for the best experience.
The free tier becoming the ad tier is how that contradiction resolves.
Notice the opt-out option they're testing: fewer free messages in exchange for no ads. That's a preview of where this goes. Ads become the default. Opting out comes with costs.
Our read: This is OpenAI doing exactly what Anthropic predicted. The test will "succeed," the rollout will expand, and the principles will quietly erode. Watch the Enterprise tier; that's where the pressure eventually lands. If business customers start seeing sponsored recommendations in their workflows, you'll know the ratchet has turned completely.