The First Shot: How a quiet ad policy change signals a new AI platform war

Sometime in the past few weeks, OpenAI quietly updated its ad policy to bar campaigns for standalone image- and voice-generation tools across the board inside ChatGPT, not a one-off decision aimed at a single advertiser. No announcement, no policy blog post: the change surfaced when Adobe, running ads for Firefly AI and Acrobat Studio through OpenAI's original pilot, was told directly that its campaigns “would no longer be approved.”

The timing is not incidental. OpenAI has just shipped ChatGPT Images 2.5 and ChatGPT Live voice, its own image- and voice-generation features.

The categories it has since closed to advertisers industry-wide, as exemplified by the Adobe case, are exactly the categories it now competes in. Video-generation ads remain untouched, which is telling: this isn't a blanket ban on rival AI tools, it's a restriction that tracks OpenAI's own product roadmap as it ships.

It is also barely a fortnight since OpenAI switched on self-serve ChatGPT Ads in India, taking a business that crossed a billion-dollar annualised run rate abroad into one of the world's largest, most price-sensitive ad markets.

None of this is happening in a vacuum. ChatGPT counts more than a billion monthly users, most on the free tier, and OpenAI is reportedly chasing a $2.5-billion annualised ad-revenue target after crossing a $1-billion run rate earlier this year. It makes sense that OpenAI is also the platform moving fastest to define what advertising inside a chatbot should look like, and, in the same breath, who doesn't get to take part.

OpenAI isn't alone in building this out, just further along.

Google keeps ads out of the standalone Gemini app for now, commercialising conversational AI through Search instead, via AI Overviews and AI Mode formats like Conversational Discovery ads and Direct Offers. Microsoft is building separate commercial formats around Copilot. Perplexity tried ads and pulled back over trust concerns; Meta treats AI conversations more as an advertising signal than an ad surface; Claude has no ads at all.

Also Read: What does an ad look like inside AI? ChatGPT and Google are starting to answer

ChatGPT Ads just hit $1bn run rate. Now comes the price test in India

Nvidia just bought the AI sandbox: Why its $13 bn Hugging Face deal matters

All are still working out the same question ChatGPT is: how close a paid message can sit to an AI-generated answer before users stop trusting either.

Checks turned up no policy at Meta AI, Copilot or Gemini resembling OpenAI's rival-tool restriction. OpenAI isn't following anyone else's rulebook here, because there isn't one yet. It didn't create this tension so much as arrive at it first: being furthest ahead on ads inside a chatbot means it's also first to run into the harder questions of who gets to advertise on it.

Will everyone else do the same?

Most people watching the space think so, eventually, and not necessarily the same way.

“ChatGPT ads switched on in India in late August 2026, and since early September any business can run them for as little as ₹725 a day. That's cheap enough that even small businesses can try it,” says Saket Dandotia, Co-founder and CEO of Onetab.ai. “But here's the reality check: most Indian companies are just experimenting right now, throwing small test budgets at it to see what happens, not making it a serious part of their marketing plan. Also, some of India's biggest ad spenders (loan companies, insurance, hospitals) aren't even allowed to advertise there yet because of OpenAI's strict rules.”

“Every AI platform that's building native generation features has the exact same incentive OpenAI just acted on,” says Aditya Kathotia, Founder and CEO of Nico Digital. Meta's own image tools, Google's Nano Banana and Veo, Microsoft's Copilot Designer, all sit in the same bind. The reason none have moved as visibly yet, he argues, is that their in-assistant ad businesses are earlier-stage than ChatGPT's; once that inventory matures, expect “the same quiet carve-outs, probably announced the same way OpenAI did it: not at all, just enforced.”

For brand-side marketers, the discomfort sits elsewhere. Dhiraj Sharma, Head of Marketing and Public Relations at Panasonic Life Solutions India, frames it as a transparency problem: as AI assistants evolve into “discovery, recommendation, and commercial platforms,” the line between an organic recommendation and a paid placement is what advertisers need protected. “Measurement shifts from metrics to signals,” he says, and signals are harder to audit than a media plan.

Meher Patel, Founder of Hector, a Wondrlab company, pushes that into a test of consistency. Category restrictions are ordinary when they protect safety or user experience; the concern begins when rules are opaque or unevenly applied. OpenAI's move has been reported “without a detailed public explanation of its competitive rationale,” and the real question, for OpenAI now and every other platform eventually, is whether each holds its own products to the standard it applies to everyone else's.

The stakes are sharpest outside the US. Pankaj Srivastava, Founder and CEO of UnoSearch, points to OpenAI's own figures, over 100 million weekly ChatGPT users in India as of February 2026, to argue that exclusion from this surface isn't a rounding error for a smaller AI company; it can mean losing a live customer-acquisition channel outright. His ask is procedural: predictable, published eligibility rules.

How much of this is allowed?

That depends on a question nobody has yet answered in court: is OpenAI dominant?

Akshayy S Nanda, Partner at Saraf and Partners, is unambiguous that dominant firms are, as a rule, allowed to choose who they deal with. A refusal to deal only becomes unlawful under the “essential facility” doctrine, where a firm controls infrastructure rivals genuinely need to compete. Self-preferencing is narrower: it requires substantial market power, high entry barriers, strong network effects, and a real foreclosure effect on rivals. Dominance, Nanda stresses, is “the indispensable gateway to liability;” without it, the same conduct falls outside the reach of competition law.

Ankit Sahni, Partner at Ajay Sahni & Associates, agrees there is nothing unlawful, in principle, about a platform deciding what it will or won't advertise. “The competition law question, however, becomes more complicated where the platform is itself competing with the advertisers it excludes, and particularly where that platform has acquired sufficient market power to function as an important gateway to consumers.”

He points to a detail that sharpens the timeline: OpenAI's own advertising policy, updated on September 10, “now expressly reserves its right to decline advertising where it conflicts with its 'business interests' or 'competitive position.” Sahni flags Section 4 of India's Competition Act as particularly relevant, since it addresses both denial of market access and the use of dominance in one market to protect another.

Between them, Nanda and Sahni sketch the same threshold from two directions: dominance is what turns an ordinary business decision into a legal one, and the timing and wording Sahni points to are the kind of evidence that question would turn on.

Meanwhile as Nanda says, “Once dominance is established, the firm bears a special responsibility not to abuse that position. Refusing access to an essential input or systematically favoring its own downstream services over those of rivals may then attract scrutiny for anti-competitive effects.”

What is already clear is that advertising access and product competition have started to collapse into the same decision inside these platforms, and that shift, not any single company's policy, is what e4m will turn to in an upcoming story.