ChatGPT Ads 101: Advice on budget, benchmarks and barriers

For advertisers considering ChatGPT Ads, the biggest question is not whether consumers are using AI to research purchases, but whether paying to reach them there can deliver better business outcomes than established media channels.

Limited local performance benchmarks mean advertisers entering the platform must establish their own baselines and determine whether conversational advertising can generate incremental business, rather than simply redirect existing media spending.

Arya Nugraha, country director for Indonesia at M+C Saatchi Performance, recommended treating early campaigns as structured learning exercises, with measurement and commercial targets agreed before launch.

He suggested a starting budget of around US$20,000 a month, clarifying that this was the agency's guidance rather than a platform minimum.

Before spending, advertisers should establish conversion tracking, agree on reporting and define what success looks like. Nugraha also advised keeping campaign structures simple rather than spreading budgets across too many ad groups, which could lead to under-delivery.

With local benchmarks still limited, he said advertisers should compare ChatGPT Ads against their own search and social campaigns for the same products.

"You're going to be comparing the cost and quality of the leads or sales you get from your search and social campaigns for the same products," he said.

You need to write down your scaling criteria prior to launch so that the call to scale, the call to keep testing or the call to pull back is based on actual agreed numbers rather than just your gut.

Don't miss: Shopee lands in ChatGPT as Sea and OpenAI deepen partnership

Conrad Victorino, head of media and GM at DO. Asia, advocated a similarly disciplined approach for advertisers, recommending one proposition, a controlled budget and a measurable action.

He suggested an initial testing window of four to six weeks, provided campaign delivery supports it.

"Scale if results meet agreed commercial thresholds. Refine if engagement is promising but conversion needs work. Stop if sufficient testing produces neither. A pilot should end with a recommendation, not another presentation about potential," he said.

Attribution does not prove incremental growth

For performance marketers, the challenge extends beyond whether advertisements generate clicks or conversions. They must also determine whether those outcomes would have happened without ChatGPT Ads.

Nugraha said click-through rates and cost per click could help identify campaign problems, but advertisers should prioritise lead quality, conversion rates, cost per acquisition and, where possible, customer lifetime value.

He pointed to available measurement tools including a pixel, conversions API and UTM tracking, alongside click-through and view-through attribution. Mobile measurement partners AppsFlyer, Adjust, Branch and Singular are also supported, he said.

Even with these tools, attribution alone cannot establish whether ChatGPT Ads generated additional sales. A consumer might encounter a brand through ChatGPT but complete the purchase through Google Search, leaving the contribution of the earlier interaction unclear.

"On incrementality we know attribution on its own won't prove it," Nugraha said.

"The only way to find out is to run controlled tests (e.g. hold back by region or by audience) and then see what happens to branded search and direct traffic. That's how we'll measure whether the conversation actually made a difference."

Victorino recommended comparing similar markets, running ChatGPT Ads in one while keeping other marketing activity broadly consistent in the other. Advertisers could then assess whether qualified leads or sales improved more in the test market, taking previous performance trends into account.

Where controlled comparisons are not possible, he advised agencies to report attributed results without presenting them as proof of incremental growth.

Trust and local relevance could determine adoption

Beyond measurement, ChatGPT Ads must overcome practical barriers before advertisers can incorporate it into their media strategies at scale.

Nugraha identified measurement confidence as a particular concern in Indonesia, where performance-led advertisers are accustomed to mature reporting dashboards. He also highlighted the challenge of matching advertisements to local-language conversations, given that consumers may switch between Bahasa Indonesia and English within a single sentence.

Integration is another consideration. Brands and agencies already have established teams and reporting processes for Google, Meta and TikTok, meaning a new platform must justify the additional operational work. Nugraha pointed to the Advertiser API as one tool that could help with integration.

Brand safety and privacy will also shape adoption.

Nugraha said OpenAI's decision to place advertisements below answers rather than within them could help preserve trust, but advertisers would still need transparency around ad matching and sufficient data to evaluate performance.

Victorino said OpenAI would need to provide clear eligibility rules, relevant ad delivery, safeguards around unsuitable conversational contexts and reporting that connects to business outcomes.

People come to ChatGPT for help. Advertising has to respect that relationship if brands want a lasting opportunity there.

Ultimately, the platform's ability to attract advertisers will depend on more than access to a new audience or the promise of more contextual targeting. Agencies need to show that the channel can deliver qualified customers, demonstrate incremental value and integrate with existing campaigns.

For advertisers, early adoption may offer an opportunity to learn before competition intensifies. But without agreed benchmarks and credible evidence of commercial returns, curiosity about AI advertising will not be enough to secure a lasting share of the media budget.

This is a two-part analysis. Read if ChatGPT can disrupt the media mix story here.

Related articles:
When AI gets too personal: Why ChatGPT is suddenly dropping f-bombs
OpenAI gives teens their own ChatGPT experience
OpenAI reportedly plans ChatGPT ‘superapp’ overhaul