Australian businesses have been able to buy ChatGPT Ads since May 2026, the self-serve Ads Manager is open, and there is no minimum spend. The honest position after four months of public data is that this is a real auction with real inventory and genuinely weak published performance: the two documented advertiser tests published so far recorded click-through rates of 0.65% and 1.30%, average costs per click between CA$7.16 and US$5, and almost no completed conversions. That does not make it worthless. It makes it a pilot line item with a defined kill date, not a channel you move Google Ads budget into.
Quick reference:
- Live in Australia since roughly February 2026 for users, with self-serve advertiser access from May 2026
- Ads only reach Free and Go tier users — Plus, Pro, Business, Enterprise, Edu and under-18 accounts see nothing
- Fewer than 20% of eligible users saw an ad on a given day during OpenAI’s pilots
- OpenAI’s recommended opening bid is US$3–5 CPC, roughly $4.60–7.70 AUD
- Geographic targeting is country-level only outside the United States, with no state, metro or postcode control
- No minimum spend — the original US$50,000 floor was dropped
- Health, mental health and political advertisers are restricted in Australia
What are ChatGPT Ads, and who actually sees them?
ChatGPT Ads are sponsored placements that appear beneath or inside a ChatGPT answer when the conversation context matches an advertiser’s offer. They are labelled as sponsored and visually separated from the organic response, and the format reads as native to the thread rather than as a banner. OpenAI set out the format and the safeguards in its announcement on testing ads in ChatGPT.
The single most important constraint is who is eligible to see them. Ads serve only to logged-in adult users on the Free and Go plans. Anyone on Plus, Pro, Business, Enterprise or Edu sees no advertising at all, and accounts OpenAI judges to belong to under-18s are excluded.
That cuts the addressable audience in a very specific direction. ChatGPT passed one billion weekly active users in August 2026, and around 45% of Australians have tried generative AI with roughly 28% using it weekly, so raw reach is not the problem. The problem is composition. The people most likely to be paying for Plus or Pro are the heavy commercial users: consultants, developers, marketers, founders. If those are your buyers, the ad-eligible half of the audience is the wrong half.
How do you buy ChatGPT Ads in Australia?
Two routes exist. Agency and technology partners including Dentsu, Omnicom, Publicis, WPP, Adobe, Criteo and StackAdapt have had access since the pilot phase, and Australian brands were part of that early group from around April 2026. The second route, which is the one most businesses will use, is the self-serve Ads Manager that opened to Australian advertisers in May 2026, as SmartCompany reported at the time.
Setup runs roughly as follows: create the advertiser account, complete verification, add billing, then build a campaign by choosing an objective, dates, daily budget, locations and platforms. Below that sit ad groups, which carry your bid and your context hints, and below those the creative.
Creative specs are tight. You get a title of up to 50 characters, body copy of up to 100 characters, an image, a favicon and a target URL. That is closer to a responsive search ad headline than to a Meta creative, and it means your landing page carries almost all of the persuasion load. If yours has not been through a landing page review, do that before you spend anything here.
One operational detail catches people out: you must allow the OAI-AdsBot and OAI-SearchBot user agents at your edge. Aggressive bot management on Cloudflare or Akamai will silently fail your ad reviews, and OpenAI has not published an IP range list yet, so user-agent matching is the only reliable approach.
What do ChatGPT Ads cost in Australia?
There are three objectives, and they price differently.
Reach is priced on CPM, with a default maximum bid of US$60 per thousand impressions and reported market rates between US$25 and US$60. In Australian dollars that is roughly $39 to $93 per thousand impressions, which is expensive for undifferentiated awareness.
Clicks is priced on CPC. OpenAI recommends opening at US$3–5, about $4.60–7.70 AUD. Real observed costs have run higher: Choice OMG’s June test averaged CA$7.16 per click, which is around $8 AUD.
Conversions uses oCPC, and this is the one people misread. You set a bid cap expressed as the maximum you will pay for a conversion, but OpenAI still bills you per valid click. The bid cap is an optimisation input, not a price guarantee. You can spend a full daily budget on clicks and record zero conversions, and nothing about that is a billing error.
All three run through a relevance-weighted second-price auction, so your bid is only one input. Ad and landing page relevance to the conversation also determine whether you show, which is the same underlying logic as Quality Score in Google Ads. One constraint worth planning around: campaign-total budget is a spending limit, not a pacing control, so the daily budget is your only real throttle.
Do ChatGPT Ads actually work?
This is where the marketing and the evidence diverge, so it is worth being precise about what is actually known.
The platform is commercially successful for OpenAI. By OpenAI’s own account, ChatGPT Ads reached a US$1 billion annualised revenue run rate in under 200 days and is used by tens of thousands of advertisers. That tells you money is going in. It does not tell you what comes out.
The advertiser-side evidence is thin and unflattering. SE Ranking ran more than 97,000 impressions and recorded a 1.30% click-through rate with, in their words, very few sign-ups. Choice OMG’s June test recorded 8,940 impressions, 58 clicks, a 0.65% click-through rate, an average CA$7.16 cost per click, and zero leads submitted. OpenAI itself states it does not yet have vertical performance benchmarks to share.
There is one genuinely positive signal. A Criteo retailer sample found that traffic referred from large language models converted at roughly 1.5 times the rate of other referral channels. That is consistent with what we see in analytics across the e-commerce accounts we manage: AI-referred sessions are small in volume but unusually well qualified, because the model has already done the comparison work before the click.
Read together, the picture is a channel with high intent per click, low volume, high cost per click, and no proven path from click to conversion at scale yet. That is a normal profile for a platform six months old. It is not a normal profile for a channel you fund from your primary budget.
Why does country-level targeting rule out most local businesses?
This is the detail that decides the question for a large share of Australian advertisers, and it is barely discussed.
Outside the United States, geographic targeting is country-level only. US advertisers can target by state, DMA and ZIP code. Australian advertisers can target Australia. There is no state, no metro, no radius and no postcode control.
If you are a plumber in Adelaide, a dental practice in Brisbane or a conveyancer in Perth, that means every impression you pay for is served nationally, and you are funding clicks from users who can never become customers. No amount of context hinting fixes a geography problem, because context hints are broad thematic guidance written at ad group level, not keywords and not locations.
The other targeting gaps compound it. There is no demographic targeting, no lookalike audiences, no site retargeting and no language selection. Custom Audiences exist in Australia, uploaded as email or phone lists, but require a minimum of 25,000 matched users, which most Australian SMBs simply do not have.
The practical rule: if your service area is smaller than the country, ChatGPT Ads is not currently a channel for you. Put that budget into Local Services Ads or geo-targeted Search instead.
How does conversion tracking work?
Measurement is more mature than the targeting, and it needs to be live before you launch, because the conversions objective is not available until tracking is running.
The OAIQ JavaScript pixel sets a first-party cookie named __oppref with a 30-day lifetime and supports page, content, commerce, lead, subscription and custom event types. Advanced matching is on by default and hashes form data with SHA-256 in the browser.
The Conversions API is the server-to-server alternative, and OpenAI states plainly that it is more reliable than the pixel alone. It accepts events from web, app, offline, in-store, phone and email sources, and deduplicates against browser events using shared event IDs. If you have already built server-side tracking through GTM, you have most of the infrastructure. If your conversion tracking is not verified accurate today, fix that before adding another platform to the mess.
Two reporting quirks to plan around. Conversions take 24 to 48 hours to appear, so do not judge a day on the same day. And view-through conversions report separately as VTA 1d, and explicitly do not affect your CPA, post-click conversion rate or bidding. Anyone quoting you a ChatGPT Ads CPA that includes view-throughs is quoting a different number to the one the platform optimises against.
Who should pilot ChatGPT Ads, and who should not?
Worth a pilot now. National e-commerce brands with broad consumer products, where country-level targeting is not a handicap. Considered-purchase categories where people genuinely research in ChatGPT before buying: travel, home improvement, insurance, software comparison, education. Advertisers already at their ceiling on Google Ads who need incremental reach and can fund learning. Brands where being early to a channel has strategic value beyond the immediate return.
Not yet. Local and service-area businesses, for the geography reason above. B2B companies, because your buyers sit disproportionately on paid ad-free tiers, which is a structural problem rather than an optimisation one, and the Google Ads and LinkedIn combination remains the better use of B2B budget. Health, mental health and political advertisers, who are restricted, and adjacent health categories where content safety filters cut available inventory hard. Any account spending under about $3,000 AUD per month, where every dollar should be buying proven intent.
The general test is the same one we apply to any new channel: does this reach demand you cannot already reach more cheaply? For most Australian accounts we audit, the answer in September 2026 is still no.
How do you run a six-week pilot without wasting the budget?
If you are in the “worth a pilot” group, run it as a structured test with a defined end date, not as a new always-on line.
Cap the exposure. Between $1,500 and $3,000 AUD total across six weeks is enough to learn something and small enough that a zero result does not hurt. Use daily budgets, since the campaign total is a limit rather than a pacing control.
Get tracking live first. Pixel and Conversions API both, with deduplication configured. Without it you cannot use the conversions objective and you cannot measure the outcome.
Start on Clicks, not Reach. CPM buying at US$25–60 tells you nothing actionable at this budget. Bid at the low end of the recommended range and let relevance do the work.
Write a kill criterion before you launch. Ours is: if six weeks produce fewer than five tracked conversions at any cost, or a cost per acquisition more than three times your Google Ads blended CPA, stop. Deciding this in advance is what stops a pilot quietly becoming a permanent line item.
Measure incrementality, not channel totals. Watch whether total business conversions moved, not whether the ChatGPT campaign reports conversions. This is the same discipline that exposes Performance Max cannibalisation, and it applies identically here.
Running a ChatGPT Ads pilot with us
We set up and run ChatGPT Ads pilots for Australian businesses as a fixed-scope engagement rather than an ongoing retainer, because the data does not yet justify one.
The scope is deliberately narrow: advertiser account setup and verification, OAIQ pixel and Conversions API implementation with deduplication, edge configuration so OAI-AdsBot is not blocked, campaign and creative build within the 50 and 100 character limits, and six weeks of management with weekly reporting measured against your existing Google Ads performance on the same conversion action.
What you get at the end is a documented decision: scale, hold, or stop. If the answer is stop, we say so and tell you where that budget performs better. We would rather lose a line item than bill you for a channel that is not returning, and given the published data, stop is currently the most likely outcome for most accounts. If you want the same rigour applied to the channel that is already paying your bills, that is a Google Ads audit instead, and it is usually the better first spend.
Frequently asked questions
Are ChatGPT Ads available in Australia right now? Yes. Ads have been serving to Australian ChatGPT Free and Go users since early 2026, and the self-serve Ads Manager has been open to Australian advertisers since May 2026. You can also buy through agency partners including Dentsu, Omnicom, Publicis and WPP.
How much do ChatGPT Ads cost? There is no minimum spend since OpenAI dropped its original US$50,000 floor. OpenAI recommends an opening bid of US$3–5 per click, roughly $4.60–7.70 AUD, and reported CPM rates run US$25–60 per thousand impressions. Observed costs in published tests have been higher than the recommendation, averaging around $8 AUD per click.
Can I target a specific Australian city or state? No. Geographic targeting is country-level only outside the United States, so Australian advertisers can target Australia and nothing narrower. There is no state, metro, radius or postcode option. This is the main reason local and service-area businesses should not use the platform yet.
Do ChatGPT Ads convert? The published evidence is weak. Two documented advertiser tests recorded click-through rates of 0.65% and 1.30% with almost no completed conversions between them, and OpenAI states it does not yet have vertical benchmarks. The counterweight is a Criteo finding that LLM-referred traffic converts at roughly 1.5 times the rate of other referral channels, so intent quality per click appears genuinely high even though volume and proven outcomes are not there yet.
Should I move budget from Google Ads to ChatGPT Ads? No, not in September 2026. Google Ads reaches buyers at a known cost with mature targeting and measurement; ChatGPT Ads reaches a subset of Free and Go tier users with country-level targeting and no vertical benchmarks. Fund a ChatGPT pilot from a test budget, and if you are unsure what your channels can afford, our budget calculator will give you the maths.
Is advertising in ChatGPT the same as ranking in ChatGPT answers? No, and the distinction matters. Ads are paid, labelled placements that sit beside the answer. Being cited inside the organic answer is earned, and it is a content and structure problem rather than a media buying one. That discipline is generative engine optimisation, and for most businesses it is the higher-return investment of the two right now.
ChatGPT Ads is six months old in Australia, priced like a mature channel, and measured like a beta. That combination is exactly when honest testing pays and enthusiastic spending does not. We will tell you whether your account is one of the few that should pilot it, and if it is not, we will tell you that instead. Get in touch and we will look at your numbers first.