The assistant is becoming the ad-supported action layer

ChatGPT Go users in India are about to sit in an odd middle band: distinct from the free product, but still close enough to the mass-market surface to see ads. TechCrunch reported that OpenAI will start showing ads to ChatGPT Free and Go users in India.

The obvious read is simple: OpenAI wants more revenue than subscriptions and enterprise deals can provide. True, but too narrow. The more interesting shift is that the assistant is becoming the place where commercial intent is formed, priced and routed.

Search monetised the query. Social monetised attention. Assistants are starting to monetise the moment before action.

That is a different object.

When someone asks a chatbot for advice on a phone, a course, a hotel or a flight, they are not browsing in the old sense. They are compressing research, preference and decision into one conversational exchange. If ads enter that surface, the ad unit is no longer sitting beside information. It is competing to shape the answer that becomes the next step.

Google’s travel updates make the same point from the opposite direction. TechCrunch reported that Google’s AI Mode can track flight prices and help book hotels. This is not the old search results page with a smarter paragraph at the top. It is conversational search turning into trip planning, then into transaction routing.

That distinction matters for builders. In a classic funnel, you could argue over attribution: the ad introduced the customer, the comparison page educated them, checkout completed the sale. In an assistant interface, those steps collapse. The same surface can understand the user’s intent, narrow the options, recommend a provider and push the user towards booking. Whoever owns that surface gets to decide what counts as a good answer, a relevant ad, a trusted partner or a completed task.

Gemini Live pushes the pattern beyond travel and ads into everyday work. Google described new productivity features in Gemini Live as a way to turn voice into action. Voice is being recast as a control surface, not a novelty input method.

That matters because voice lowers the friction between thought and delegation. “Remind me”, “summarise this”, “book that” and “find me a cheaper option” are all small expressions of intent. Once the assistant can act on them, the interface becomes a market maker between the user and the services that want to fulfil the request.

The useful parallel is supermarket slotting fees. Brands did not only compete on product quality; they paid for placement at the point where shoppers made decisions. Eye-level shelf space had economic power because it sat inches from purchase. Assistants may become the digital version of that shelf, except the shelf talks back, knows context and can steer the next action.

This is why “ads in ChatGPT” is too small a frame. Ads are the first visible business model for a much larger control point: the action layer. OpenAI is testing it through ads in a high-growth market. Google is pushing it through AI Mode transactions and Gemini Live productivity. Different products, same direction.

For product teams, the question is uncomfortable. If users increasingly express intent inside assistants, your app may not be the destination. It may be a capability the assistant calls, ranks or ignores.

The next platform fight will not be over who gives the best answer. It will be over who owns the second after the answer, when the user says: “Do it.”

---