ChatGPT Ads in Australia: what the first three months actually tell us

What changed, and when

For most of the last two years, being visible inside ChatGPT meant one thing: getting your brand mentioned in the answer itself. That is still the prize. But since April there is a second way to show up, a paid one, and it is now open to any Australian business with a credit card.

OpenAI began testing ads in the United States in February 2026, then moved faster than almost anyone expected. Australia went live on 17 April, one of the first markets outside the US alongside New Zealand and Canada. OpenAI pointed to Australian usage doubling over the prior year and a local team on the ground as the reasons we were early (reported by Shout Digital, citing OpenAI).

Two months later the buying side grew up. By late May the self-serve Ads Manager was open with no minimum spend, cost-per-click bidding had arrived, and OpenAI had shipped its own conversion tracking (Marketing-Interactive, May 2026). A channel that started as a seven-figure experiment for holding-company agencies became something a suburb accountant could switch on in an afternoon.

Who you are actually reaching

The single most important fact about ChatGPT ads gets missed in most write-ups: they only show to people on the Free and Go tiers. Anyone paying for Plus, Pro, Business, Enterprise or Education sees nothing (OpenAI). The heaviest, most commercially valuable users, the ones running work through ChatGPT all day, have effectively opted out of your ad.

So the ad-eligible audience leans toward casual and early-stage users. That is not a dealbreaker. It shapes strategy. You are reaching people near the top of a decision, exploring a category or comparing options, rather than someone with their card out. Expect to seed consideration, not to close a sale on the first click.

People are often learning about a category, comparing options, or deciding what to do next. In those moments, a click can be a meaningful signal.

OpenAI, on why it added CPC bidding

What the performance numbers look like

Nobody should pretend this is a mature channel. Click-through rates reported in the early pilot sat around 0.9%, a long way short of the rates a tight Google Search campaign pulls on high-intent terms. Impression volumes, on the other hand, climbed fast once inventory opened up.

The more useful signal is what happens after the click. Criteo reported that visitors arriving from large language model referrals convert at roughly 1.5 times the rate of other digital channels, and tend to spend noticeably longer on-site (reported industry data). People who click through from a conversation have usually done some thinking first. They arrive warmer.

The catch is attribution, and it is a big one. Australian agencies running early campaigns report that around 60% of the conversions ChatGPT ads influence do not happen on that first visit. They arrive later, as a branded search or a direct visit, once the person has gone away and thought about it (Multimediax, June 2026). If you judge the channel only on last-click, you will almost certainly undercount it and switch it off too soon.

How the targeting really works

Forget keyword match types. ChatGPT ads do not run on exact keywords the way Google Search does. You give the platform Context Hints, plain-language descriptions of the conversations where your ad belongs, and OpenAI decides when a chat is relevant enough to serve it (Louder, June 2026).

That means you are handing more of the targeting decision to the model than you would on any other platform. It also means the old skill of mining a search-terms report and bolting on negatives does not fully apply yet. Campaign objectives are still basic too: at launch you optimise for Views or Clicks, with conversion-optimised campaigns rolling out gradually and not yet available to every advertiser.

Practically, the businesses getting sensible results are the ones writing genuinely useful ad copy for a specific moment in a decision, rather than lifting a Google headline and hoping. The ad sits inside a conversation. It has to read like it belongs there.

The mistake most advertisers are making

The costly error is not the money spent on clicks. It is running the channel blind. Because so much of the payoff lands later and off-platform, campaigns that look mediocre inside OpenAI’s own dashboard are often doing more than the numbers show, and campaigns that look fine can be leaking budget on junk clicks nobody is checking.

• Install the measurement properly. Use both the browser pixel and the server-side Conversions API. The server-side route survives ad blockers and cookie limits, so it captures conversions the pixel alone loses.
• Watch branded search and direct. Track those channels for lift during a campaign, not just the ChatGPT numbers, or you will miss most of the impact.
• Set a learning budget, not a bet-the-farm one. This is an experiment with real upside, funded from test budget, not money pulled out of a proven Google account.
• Keep organic AI visibility first. Being recommended inside the answer beats a sponsored box beneath it, and OpenAI has confirmed ads never influence those organic recommendations.

// Key takeaways

• ChatGPT ads reach a large but casual, early-stage Australian audience, paid-tier users never see them.
• Click-through rates are low by Google standards, but LLM-referred visitors convert better once they land.
• Around 60% of the impact shows up later as branded search or direct, so last-click reporting will undersell it.
• Targeting runs on Context Hints, not keywords, so ad copy and conversation-fit matter more than match types.
• The channel is worth testing with a learning budget and real tracking, while organic AI visibility stays the priority.