Surprisingly Useful AI Article Enhancements
A Sydney digital marketing agency is warning that Australian businesses adopting ChatGPT advertising risk drawing the wrong conclusions about whether the channel works, because standard tracking captures only a fraction of the conversions it generates.
Multimediax, a Google Premier Partner agency founded in 2001, has flagged what it describes as an attribution gap unique to AI-driven advertising. According to the agency’s analysis of early campaign data, only around 40 per cent of conversions generated by ChatGPT ads can be tracked directly to the ad itself. The remaining 60 per cent arrive later, when a person who encountered an ad mid-conversation returns days or weeks afterwards through a branded Google search or by visiting the advertiser’s website directly.
Under a standard attribution model, that delayed 60 per cent is credited to other channels or to nothing at all.
“Everyone’s talking about how to get into ChatGPT ads. Not many are talking about how to measure them, and that’s where the money gets lost,” Najib Raad, Managing Director of Multimediax, warned. “If you judge this channel the way you judge Google Ads, the numbers will lie to you. You’ll see 40 per cent of the picture and make a decision as if it’s the whole thing.”
How ChatGPT ads reached Australia
ChatGPT began showing ads to Australian users in April 2026, making Australia one of the first markets in the world to receive them alongside Canada and New Zealand. The rollout moved from a closed experiment requiring around $200,000 in committed spend to a self-serve platform with no minimum budget in a matter of months.
OpenAI launched ads in the US in February 2026 with a small group of large advertisers. The pilot crossed $100 million in annualised revenue within roughly six weeks, with more than 600 advertisers signed on. The self-serve Ads Manager opened in May 2026, bringing CPC bidding and OpenAI’s own conversion tracking with it. Conversion-optimised campaigns began rolling out in June 2026, though they are not yet available to all advertisers.
Raad argued the speed of the rollout is precisely why caution matters.
“This went from locked behind a $200,000 spend to something you can set up quite quickly. That’s great, but it also means a lot of businesses are launching with no measurement plan and high expectations,” he told journalists.
A different targeting model
Unlike Google Ads, ChatGPT ads do not use keyword bidding. Advertisers instead write what OpenAI calls “context hints” – short, plain-language descriptions of the conversational situations where their product is relevant.
Rather than bidding on a term like “project management software,” an advertiser might write a hint such as “someone comparing CRM tools for a small sales team” or “considering switching away from a tool they find too complicated.” The system matches ads to the meaning of a conversation rather than to specific search terms.
Ads appear in a clearly labelled sponsored box beneath ChatGPT’s response. OpenAI has maintained that ads do not alter the AI-generated answer above them, that the model generating the answer is not aware the ad is present and that advertisers never see anyone’s conversation content.
The audience is also narrower than the full ChatGPT user base. Ads are shown only to users on the free plan and the lower-cost Go plan. Users paying for Plus, Pro, Business, Enterprise or Education tiers do not see ads at all, which means the advertising audience skews towards casual and early-stage users.
The attribution gap
The measurement challenge is the central concern Multimediax has raised. People engaging with ChatGPT ads tend not to click through in the same session. They encounter an ad while working through a decision in conversation, then return to the advertiser later through a separate search or direct visit.
Multimediax reports that early advertiser data puts the size of this gap at roughly 60 per cent. In a standard attribution model, none of those delayed conversions are credited to the ChatGPT campaign.
“We’re seeing it in our own campaigns. The direct conversions look modest, but branded search lifts while the ads are running and softens when we pause them. That’s the channel working,” Raad explained.
Related: Best Business Laptops for work & school
Related: Best Gaming Laptops
Related: Best Portable Laptop
The agency recommends businesses track branded search volume and direct website traffic during campaign flights to identify whether those numbers rise when ChatGPT ads are active and fall when they are paused. OpenAI offers both a browser-side tracking pixel and a server-side Conversions API. The server-side option is more reliable, as it survives ad blockers and cookie restrictions.
Reporting is aggregated by design. Advertisers will not see individual user data or conversation content, making the measurement model closer to connected TV than to traditional search.
Which businesses should consider the channel
Multimediax’s assessment is that ChatGPT ads suit businesses where customers research before purchasing. The agency points to B2B and SaaS companies, where buyers compare vendors before engaging sales teams, as well as higher-value retail categories such as furniture and appliances where deliberation is typical.
Service businesses with longer sales cycles and education and training providers are also identified as categories likely to benefit, given that users frequently turn to ChatGPT to compare options in those areas.
The channel is less suited to impulse purchases or businesses that depend on tight geographic targeting. Targeting in Australia is currently limited to the country level, with no suburb or city-level options available. Local businesses paying to reach a national audience would see limited return.
Whole industry categories are also barred from the platform. Health and medical, legal, finance, gambling, alcohol and tobacco, adult content and political advertising are all excluded – a list broader than Google’s own restrictions.
A realistic test requires a daily budget of roughly $50 to $100 over four weeks, according to the agency. Below that threshold, campaigns do not generate enough data for the system to optimise effectively. When CPC bidding launched, OpenAI suggested a starting range of $3 to $5 per click, though reliable Australian cost-per-lead benchmarks do not yet exist given how new the channel is.
Not a replacement for Google
Multimediax has cautioned against treating ChatGPT ads as a substitute for existing search advertising.
“Google catches someone the second they decide to search and are ready to act. ChatGPT catches them earlier, while they are still figuring out what to even ask,” Raad observed. “Different stage, different job.”
The agency’s position is that ChatGPT ads work as a new layer in an existing search strategy rather than as a replacement for Google Ads or paid social. The two channels reach people at different points in the decision-making process and should be measured together to understand how they influence each other.
Paid ChatGPT ads place a business in the sponsored box beneath an AI response. Getting a brand cited within the response itself is a separate discipline, falling under what the industry refers to as AI engine optimisation or generative engine optimisation. The agency’s analysis of the ChatGPT ads attribution gap and details of its ChatGPT Ads management service are available on its website.
What the agency recommends before spending
Multimediax has outlined a set of steps for businesses considering ChatGPT ads. The agency recommends installing conversion tracking before launching any campaign so that early data is not lost and the system has conversion events to optimise towards.
Businesses should monitor branded search and direct traffic for movement during campaign periods, allow several weeks before evaluating performance to account for delayed conversions and confirm their industry is eligible before committing budget.
The platform remains in beta and continues to add features. The agency has urged businesses to treat early results with appropriate caution, given the pace of change and the absence of established benchmarks in the Australian market.
Last Updated on June 28, 2026 by Nick Ross
