OpenAI switched on ChatGPT Ads in India today, starting with roughly 50 brands routed through WPP and Omnicom. On September 4, the ChatGPT Ads Manager opens for self-serve buying with a daily minimum of ₹725, about $7.60. That’s the whole story for anyone selling into India: a week of managed-agency staging, then the door opens to everyone.

The market context is the reason to pay attention. ChatGPT counted 330 million unique monthly users in India as of May 2026, up from 297 million the previous December. India is one of OpenAI’s largest and most active markets, and the company is chasing a $2.5 billion advertising revenue target for the year. The auction floor at launch is a customer-acquisition subsidy, not a permanent number.

Ads appear only on the Free tier and on Go, the ₹399-a-month ($4) plan. Plus, Pro, Business, Enterprise, and Education stay clean. Placement sits below the response, visually separated, clearly labeled as sponsored. OpenAI told CNBC the format is designed so that “ads will be clearly labeled, visually separate and shown below the response,” adding: “Answer independence is non-negotiable.”

That framing matters more than it looks. The whole product depends on users trusting that the answer above the ad wasn’t shaped by the ad below it. If that trust cracks, the inventory loses its premium and the auction collapses toward banner economics. OpenAI knows this, which is why the language is absolute rather than hedged.

Dave Dugan, head of global ads solutions at OpenAI, put the pitch this way: “With ChatGPT Ads, businesses of every size can introduce themselves at relevant, high-context moments when decisions are beginning to take shape.” Translated: this is intent inventory, closer to search than to social, but with the query expressed as a full paragraph of context rather than three keywords.

The rollout also fits an obvious sequence. It follows the day-one European go-live and the 31-market European launch earlier this month. Managed partners first, self-serve within days, low floor, large user base. Those floors don’t stay.

The September 4 window is when the interesting question gets answered: whether contextual placement inside a reasoning session actually converts, or whether it just looks like it should.

Sources