Paula Lopes, Digital Partner at Avenue C
At Avenue C, we have a rule: new channels earn their place in our clients' plans through evidence, not enthusiasm.
Our clients deserve an agency that forms views from first-hand experience and actual data, not published commentary or buzz. That's why, when OpenAI invited us to join the first beta cohort of their Ads Platform in Australia alongside the holding companies, we said yes, but selectively.
That invitation matters. Independent agencies are rarely in the room at this stage. OpenAI made a deliberate choice to include indies, and that says something about how they are thinking about the market.
We onboarded two clients who were commercially bold and genuinely committed to a test-and-learn approach. Two months in, here is what we have found.
1. The untouched audience opportunity is real
ChatGPT reaches people who have effectively opted out of traditional advertising. Digitally native, ad-blocking, disengaged from social feeds. These are not easy people to reach. Across our clients, 95% of traffic generated through ChatGPT was new users, compared to a lower average across other paid channels.
That is a significant signal.
The context is also different. They are in a decision-making frame. The intent quality is high, and for the right category, that matters.
The challenge is what happens next. Our Chat GPT audience is currently spending 31% of the average session duration compared to other paid channels. Reaching them is one thing. Earning their attention beyond the click is the next problem to solve, and that is partly a creative and landing experience challenge, not just a platform one.
2. The platform launched to get brands in early. That was the right call.
The ads manager is simple. It does not mirror the depth of Meta or Google, and it does not yet connect to ChatGPT's own capabilities to enrich data or analysis.
It reads as a minimum viable product, and that is not a criticism. OpenAI made a deliberate decision to get a working product in market so brands could participate early and build familiarity with the format. That is a reasonable trade-off between speed and sophistication, and it is one we understand well in this industry.
The pace of development backs this up. Weekly updates are rolling out in real time. Pixel implementation, bulk editing, expanded country availability, and billing changes are all being added on the fly. This is a platform that is genuinely evolving, and being in the beta means being part of that build.
3. Measurement is the gap that matters most
Campaigns can currently only be optimised towards CPM or CPC. There is no lead or conversion objective available yet, no brand lift measurement, and no audience breakdown. Right now, we are working with impressions, clicks, cost, CTR, and CPM. That is it.
This is the most important limitation to address, and not just for campaign optimisation.
It connects to a broader question the industry is actively working through: where do LLMs sit in the consumer journey, and how do we measure their contribution to business outcomes?
Consider this: 65% of searches no longer end in a traditional click. That shifts where intent lives.
Paid search still plays a role in capturing demand, but LLMs are increasingly where consideration and discovery happen, particularly in certain categories. If that is true, then LLM ads cannot be a bolt-on tactic evaluated in isolation.
They need to be planned as part of an omnichannel strategy from the start, with measurement frameworks built around the full journey, not just last-click signals.
The MMM question sits here too. Marketing mix models run on aggregated spend and outcome data. In theory, ChatGPT ad spend could be ingested like any other channel. In practice, the data export is too thin.
Without conversion signals or outcome-linked metrics, there is nothing for a model to connect to a business result. That is not a permanent state, but it is the current one, and it should inform how we plan and report on this channel in the interim.
4. The future of this platform is genuinely exciting
Imagine a report that tells you exactly which prompt generated your impression, your click, your sale, and your revenue. Imagine understanding what question a user was asking, and what organic answer surrounded your ad when it drove a result.
That kind of prompt-level insight would be an amazing evolution of what is currently available on other channels (specially search engines), and it is where this platform could logically go.
The infrastructure to get there exists. The data model is already intent-first. The question is how quickly OpenAI builds the reporting and optimisation layer on top of it.
5. This is just the beginning
We are at the start of something. ChatGPT ads are not a finished product, and they are not ready to carry commercial accountability on their own. But the audience access is real, the intent context is different from anything else in the paid media mix, and the direction of the platform is the right one.
That said, everything above could be out of date by the time you finish reading it. We genuinely hope it is.