ChatGPT Ads Malaysia: No Targeting, and That's the Point

ChatGPT Ads began rolling out in Malaysia on Wednesday 24 September 2026, alongside Indonesia, the Philippines, Singapore, Thailand, Vietnam and Taiwan, with self-serve buying available from day one (OpenAI, via TechNode Global). Most of the commentary landing in Malaysian marketing groups this week says the same thing: this is the Facebook ads moment of 2016, cheap and precise. Neither half of that is true, and the way it is untrue is the most useful thing an owner can know before spending a ringgit.

What are ChatGPT ads, and who actually sees them?

A ChatGPT ad is a single card that appears below an answer, clearly labelled as sponsored and kept visually separate from the response itself. The unit carries an advertiser name, a logo, a title, a description, one square image and a landing-page link. There is no video and no expandable format in standard campaigns, though product-feed campaigns have been able to run a multi-product carousel since August 2026 (Index Lab creative specs, updated 7 September 2026).

Ads appear to a narrower group than people assume. OpenAI's own wording is that ads are shown to eligible users on the Free and Go plans, while Plus, Pro, Business, Enterprise and Edu plans stay ad-free. So you are not reaching everyone using ChatGPT in Malaysia. You are reaching the ones who have not paid to remove ads.

What makes the inventory interesting is not its size but its moment. A Facebook ad interrupts someone looking at a friend's wedding photos. A ChatGPT ad appears while someone is describing their actual problem in their own words, usually in more detail than they would ever type into Google. Someone comparing options at 11pm has already told the system their budget, their area and their constraint before your ad ever renders.

The platform is not a pilot any more. OpenAI says ChatGPT Ads reached a US$1 billion annualised revenue run rate in under 200 days, with tens of thousands of advertisers running campaigns. Malaysia is arriving late to something already at scale, which cuts both ways: the product works, and the honeymoon pricing is gone.

What do ChatGPT ads cost in Malaysia?

Here is where the "cheap like 2016" story falls apart.

Published ChatGPT Ads benchmarks sit at roughly US$25 to US$60 per thousand impressions, averaging around US$45, with click bids clearing about US$3 to US$5 (Index Lab pricing guide, updated 24 September 2026). At the 25 September 2026 spot rate of about RM4.07 to the dollar, that is RM102 to RM244 per thousand impressions, and roughly RM12 to RM20 a click. Facebook CPMs across Malaysian accounts generally run RM8 to RM25 depending on industry and objective (ZenWeb Meta ads pricing).

Two things do make it accessible despite the rate. The US$200,000 minimum spend that launched with the platform in February 2026 was abolished when self-serve opened on 5 May 2026, and the minimum daily budget in Malaysia is RM5.50. You can genuinely test this for the price of a teh tarik a day. OpenAI also benchmarked conversion-optimised bidding at a median 44% lower cost per result than click bidding across 27 advertisers in July 2026, which is the single easiest setting to get right on day one.

How are ChatGPT ads different from Meta ads?

This is the part that will cost people money, because the difference is not a feature gap you work around. It is a different theory of how an ad finds a person.

Meta is built on the theory that who you are predicts what you will buy, so it sells you a profile. ChatGPT is built on the theory that what you are saying right now predicts what you will buy, so it sells you a moment. When the system detects commercial intent in a conversation, it runs a relevance-weighted auction among eligible ads and the most relevant ones win.

The customer-list exception deserves a straight answer, because it is the one place people will say "but you can target". OpenAI does support uploaded customer audiences, but an audience must reach at least 25,000 matched ChatGPT users before it can be used, and OpenAI recommends at least 100,000. Because people often sign into ChatGPT with a different email from the one they gave you, your raw list has to be considerably larger again. For a Malaysian agency with 3,000 past clients, the feature does not exist in any practical sense.

Frequently Asked Questions

If you cannot pick an audience, what picks it for you?

Your words do. This is the part almost nobody is writing about, and it is the whole game.

The only levers that influence which conversations your ad appears in are a title of up to 50 characters, a description of up to 100, and a list of negative phrases that block rather than attract. OpenAI recommends 16 to 24 characters for the title and 32 to 48 for the description, because on a phone anything past roughly 30 characters of title is read truncated. So realistically you are defining your entire audience in about twenty characters.

That changes two habits a Meta-trained team will carry over without noticing.

Creative testing and audience testing are now the same test. On Meta, running four titles against one saved audience is one experiment with four variants, and the audience is the controlled part. Here, four titles are four different audiences, because the title is what decides which conversations you land in. A team that reads a ChatGPT test the way it reads a Meta test will conclude that "creative three won" when what actually happened is that creative three bought a different room. Change one thing at a time, and treat every copy change as an audience change.

The words that win are the words your buyer types, not the words your industry uses. Nobody sits in ChatGPT and types "integrated interior solutions". They type "how much to renovate a 3 room condo in Bukit Jalil". Trade vocabulary in a 50-character title quietly buys you a smaller and more expensive room, because it only matches conversations held by people who already speak your industry's language. That is the opposite of what you want at the top of a funnel.

This is also why the platform rewards a specific kind of business. If you can describe what you do in the same words a confused buyer would use, you are already most of the way to a working campaign. If your positioning only makes sense after a paragraph of explanation, 50 characters is going to hurt.

How to run your first ChatGPT ads campaign from Malaysia

Conversion tracking works through either the OpenAI pixel or a server-side Conversions API, and UTM parameters on your landing URL persist through the click, so your existing analytics will see the traffic. Tag it properly on day one. In six months you will want to know what a ChatGPT-sourced lead was actually worth compared with a click-to-WhatsApp lead, and you cannot backfill that.

One operational warning, and it is the one that matters most for a Malaysian service business. A lead that arrives from ChatGPT is further down the decision than a lead that arrives from a Facebook interest audience, because they described their problem in detail before they ever saw you. They are also paying you RM12 to RM20 for the privilege of arriving. A lead that expensive and that warm sitting unanswered until the next morning is a worse loss than the same silence on a cheap cold lead, which is the whole argument behind the five-minute rule. Tools like Raion HUB's AI chatbot and agents exist to hold that first exchange, qualify in the customer's own language and assign the lead to whoever is actually on duty, so the expensive click does not die in an inbox.

Is this really the Facebook ads of 2016?

Yes, and the parallel runs backwards from how people mean it.

The popular version says 2016 was cheap. That does not survive a check. Facebook CPMs in 2016 already varied roughly US$8 to US$25 depending on client and objective (Social Fulcrum CPM data, 2016), against a 2026 average near US$11.76 (Mako Metrics benchmarks). Cost per impression is not the story of 2016 and never was.

What actually made 2016 extraordinary was two things at once. Targeting was precise to a degree that felt almost unfair, and hardly anyone was competing for the attention. You could describe a person in enough detail to make a room of 4,000 people and buy it cheap because nobody else was bidding for that room.

ChatGPT Ads in Malaysia has exactly one of those two conditions, and it is the second one. The auction here is days old. Tens of thousands of advertisers are running globally, but almost none of them are bidding on a renovation question typed in Malaysian English. The other condition is inverted: precision is not available to you at all, and the platform has taken the targeting out of your hands entirely.

So the edge available now is not "target better than everyone else". It is "write plainer than everyone else, while the room is still empty". That is a real edge, it is available to a two-person business with RM5.50 a day, and it will close the same way the 2016 one closed, as soon as enough advertisers turn up and the cost of a title in the customer's own words gets bid up.

The bottom line