ChatGPT Ads Are Now Live in Australia: Guide to Setup, Tracking & Conversions API
Conversational advertising has now reached Australia. OpenAI has rolled out its self-serve Ads Manager, with ChatGPT Ads now live in Australia, the United States, Canada, and New Zealand.
For early adopters, this is a rare window to experiment and learn while the system is still early in its lifecycle. There is upside in being first in: access to highly engaged users and less saturated inventory. But the trade-off is real, new platforms rarely deliver immediate profitability, and results can be volatile at the start. More importantly, ChatGPT isn’t just another ad channel like Google or Meta. It operates in a different environment entirely, where ad formats, targeting logic, and measurement all require a new approach.
One of the biggest risks for brands rushing in is measurement breakdown. Without correct conversion tracking in place, performance can look deceptively strong inside ChatGPT while GA4 shows messy attribution and unexplained traffic spikes elsewhere.
To train the OpenAI system effectively on your best customers, you need two things: precise contextual targeting and a solid tracking foundation. Below is a breakdown of how ChatGPT Ads work, and how to properly set up the Measurement Pixel and Conversions API (CAPI) to get reliable results.
How ChatGPT Ads Actually Work: Format & Targeting
Before attribution comes mechanics. This is not a keyword auction system, and it’s not traditional display advertising repackaged for chat.
The Format: Text + Image
Ads are delivered as native conversational inserts. A short-form text response is paired with a supporting image and surfaced directly within the chat experience when the system detects relevant intent.
The Targeting: "Context Hints" (Not Exact-Match Keywords)
This is where the model diverges sharply from Google Search.
There are no exact-match keywords, no phrase match rules, and no manual query-level control. Instead, delivery is guided by “Context Hints” high-level thematic inputs that inform when your ad should be considered relevant within an ongoing conversation.
In practice, this means targeting is semantic. The system interprets user intent in real time and decides whether your offer aligns with the conversational context.
The implication is straightforward: you are no longer controlling queries, you are influencing interpretation. That shifts optimisation away from precision targeting and toward alignment with intent signals.
The Bidding Reality: Views and Clicks Only
At this stage, optimisation is restricted to Views and Clicks. There is no conversion-based bidding objective such as “Maximise Conversions,” even though conversion tracking is supported. This is no doubt coming soon.
This creates a structural gap between spend and outcomes. You are effectively optimising for exposure and engagement, while downstream performance depends entirely on your ability to track and interpret conversions accurately.
Without robust measurement in place, you risk optimising toward superficial engagement signals while losing visibility on actual revenue impact.
The Two Pillars of ChatGPT Ads Measurement
OpenAI gives you two mechanisms to send conversion data back into Ads Manager. Treating either one in isolation is a mistake; the system is designed to work best when both browser-side and server-side signals are used together.
1. The ChatGPT Measurement Pixel (Browser-Side)
The Measurement Pixel is a JavaScript SDK (oaiq) deployed in the <head> of your website. Its role is straightforward: capture user actions at the moment they occur inside the browser.
- Privacy-preserving attribution: When a user clicks a ChatGPT ad, they arrive on your site with an
opprefparameter appended to the URL. The pixel captures this identifier and stores it in a first-party cookie (__oppref), allowing subsequent activity to be tied back to the original click without exposing user-level personal data. - Event instrumentation: Standard funnel events such as
lead_created,checkout_started, ororder_createdcan be fired, along with custom events aligned to your conversion model. - Automated handling: The SDK manages timing, batching, and source stitching, reducing the need for manual event orchestration.
2. The Conversions API (Server-Side)
The Conversions API (CAPI) exists to close the reliability gap left by the browser.
Instead of relying on client-side execution, events are transmitted directly from your server to OpenAI via HTTP POST requests, in batches of up to 1,000 events.
- Bypass of browser limitations: Because events are server-originated, they are unaffected by ad blockers or client-side restrictions.
- Extended conversion visibility: You can send downstream events that occur outside the browser session, such as CRM lifecycle changes (e.g. “Qualified Lead”), offline conversions, or phone call outcomes using
action_sourcevalues likeofflineorphone_call.
The trade-off is implementation complexity. CAPI does not automatically inherit the oppref identifier. Your system must explicitly capture it from the landing page URL, persist it in your database, and then reattach it when forwarding conversion events back to OpenAI.
In practice, this turns CAPI into an engineering-led integration rather than a plug-and-play tracking layer, but it is also what makes the dataset complete and resilient.
Don't Let Bad Tracking Ruin Great Ads
ChatGPT Ads don’t operate on deterministic keyword logic. With delivery guided by “Context Hints,” performance is heavily shaped by how effectively the system learns from your conversion signals.
That creates a hard dependency: if your tracking is inaccurate, incomplete, or misconfigured, you are effectively training the algorithm on noise. The outcome is predictable, misallocated spend, distorted optimisation, and inflated CPA/ROAS that doesn’t reflect true business performance.
Getting this right is not a superficial setup task. It involves implementing the Measurement Pixel correctly, engineering a reliable Conversions API pipeline, handling oppref capture and persistence, and ensuring server-side deduplication is correctly enforced across all events.
This is specialised implementation work, sitting at the intersection of media buying, analytics architecture, and backend engineering.
The objective is simple: ensure every conversational click is translated into a clean, usable signal for revenue-driven optimisation.
