The answer now has
a second client.

What actually changed

On 23 September, OpenAI extended ChatGPT Ads to seven markets: Malaysia, Indonesia, the Philippines, Singapore, Thailand, Vietnam and Taiwan. Ads appear only on the Free and Go plans. Go costs RM24 a month here, so even paying users at that level see them. Plus, Pro and Enterprise stay ad-free. The ads sit in labelled cards under the answer, matched to what you are asking. Advertisers buy them through OpenAI’s own sales team, the big agency holding groups and technology partners — which is where the reseller in my feed comes in.

OpenAI’s stated principles are clear: ads “do not influence the answers,” advertisers never see your chats, and the ads exist to fund free access. I will take those at their word. My argument does not depend on OpenAI breaking them.

Why an answer is different from a search result

A search page always looked like a marketplace. Ten links, some sponsored, and you did the choosing. People learned to skim past the ads.

An assistant gives you one answer, in a confident voice, in the middle of a decision. That is why people started asking it the questions that matter — the ones where they want a judgment, not a list. Its value was that it had no stake in what you chose. I saw this in our own funnel: customers talked to our chat assistant first precisely because it had nothing to sell them, and only then messaged me on WhatsApp. I wrote about it in The AI Didn’t Close the Sale. It Earned the WhatsApp Message.

The model will not get less clever when ads arrive. It will be exactly as capable tomorrow as it was yesterday. What changes is whose interest you have to wonder about. In a recommendation, neutrality is the intelligence. An answer that might be working where the money is paid is worth less, even if it happens to be the same answer.

The founders of Google said it first

In 1998, two Stanford students described a new search engine in a paper. In the appendix they wrote: “we expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers.” Their names were Sergey Brin and Larry Page.

Twenty-eight years later, everyone can judge that prediction from their own results page. The business worked beautifully. The results page filled up with things someone paid for, and a generation learned to scroll past the first screen.

Conversational ads start that cycle again, only faster, because the answer is where the decision happens. In The Landlord Just Became a Tenant. I wrote about Shopee moving inside ChatGPT. This is the next step: the discovery layer has started charging rent.

Europe has already called it a search engine

The regulators got there before the advertisers reached Malaysia. On 31 August, the European Commission designated ChatGPT’s search feature a Very Large Online Search Engine under the Digital Services Act. It described ChatGPT as “a hybrid service that qualifies as an online search engine because it responds to users’ prompts and queries, including by searching the web.” OpenAI reported about 159.1 million monthly users of ChatGPT search in the EU, more than three times the 45 million threshold.

That puts ChatGPT in the same legal category as Google Search. By January 2027 it has to assess and reduce systemic risks, and a service in that category that shows ads must keep a public archive of them. Europe is treating it as what it has become: a search engine, now with search-engine ads.

So the assistant has become the thing Brin and Page warned about, and Europe is regulating it that way. Here the ads arrived first. The public archive is an EU obligation.

To be fair: ads are not automatically a death sentence

My prediction deserves the strongest case against it.

Advertising did not destroy Google. It made Google one of the most valuable companies in history, and funded free search for billions of people who would never have paid. The same can be true here: a student in Kelantan gets a better free assistant because a hotel chain paid for a card under someone else’s travel question.

The numbers are also less one-way than the headlines. Similarweb’s traffic tracker had ChatGPT at 73.3% of generative-AI web visits a year ago and 55.5% in August. That is a big fall — but it is up from 52.7% in May. Decline is not a straight line, and a company with ChatGPT’s reach can recover.

Why I still expect OpenAI to lose ground

This is my prediction, not a certainty.

First, the product’s priorities follow the paying customer. Once the largest part of the audience is funded by advertisers, every roadmap meeting has an advertiser in the room, even an invisible one. The line between “the ad is separate” and “the answer is ad-shaped” is held by policy, and policies get revised when revenue targets are missed.

Second, competitors now have a line OpenAI cannot say. Anthropic ran a Super Bowl ad in February saying ads are coming to AI, “but not to Claude.” Disclosure: we run much of IMA on Claude, so read my view with that in mind. The point does not depend on which rival says it. Any competitor that sells subscriptions or enterprise contracts can now sell one thing OpenAI’s free tier cannot: an answer with only one client.

Third, the users most exposed are the ones who never pay. In markets like ours, that is most people. Those users will learn, as search users did, to discount the answer. An assistant whose answers are discounted is no longer the one people check first.

What I would do with this

If you buy ads: this is a real new placement, and early buyers will get it cheap. Test it with a small budget, measure hard, and do not mistake being present in the answer for being chosen by it.

If you use ChatGPT for decisions: treat the free answer like a search result. Ask the same question somewhere with no ads, or pay for a plan without them.

If you build a business on trust: notice what just became scarce. An answer with no second client is now a selling point, and it is worth protecting in your own customer conversations.

The bottom line

ChatGPT did not get less intelligent this week. It got a second client — and for a tool whose value was the judgment it gave you, that is the more expensive change. Google’s founders predicted what ads would do to search, and then built the company that proved them right. Europe now classifies ChatGPT as a search engine, and OpenAI has chosen the same path in the one place where the answer is the decision. I expect it to lose ground to rivals who can promise an answer that serves only the person asking.

If you think I am wrong about this one, I would like to hear why.