
If you’ve been hanging around the Empire9 corner of the internet for a while, you’ll know we love staying ahead of the curve when it comes to paid media, search, and all things digital domination. So, when OpenAI recently opened up ChatGPT Ads to advertisers in New Zealand, you know our performance team was already on the ground testing the waters.
What we were glad to pass along to our curious clients, is that ChatGPT ads were possible with as little as 25NZD a day, when previously, the platform looked to cost thousands for advertisers. So now you know you can afford it, your next question is “should we give it a go…?”
… and you’re not alone, we’ve been getting a stack of questions from clients about where AI placement fits alongside standard Google and Meta campaigns, so we sat down for a Q&A with Tom Rowley, Empire9’s Digital Performance Manager. Here is his take on how ChatGPT Ads work, how to set them up, and why getting in early might just be your secret weapon.
What are ChatGPT Ads, and how are they different from Google or Meta Ads?
Tom: ChatGPT Ads are reasonably new. They’ve only really become available to smaller advertisers over the last few months.
The key thing is that they show a small ad to users on free ChatGPT accounts and some lower-tier plans. Depending on the context you’ve set up in your targeting, your ad can show for different prompts people are using.
Unlike Google where you’re targeting keywords, or Meta where you’re targeting audiences, ChatGPT is much more context-based.
Who can currently run ChatGPT Ads?
Tom: They’re available to advertisers in New Zealand, and most countries can be targeted.
The setup process is pretty straightforward. You basically set up your business name, address, and complete a couple of verification steps.
From there, the account can either be managed by you directly or you can add an agency into your account to manage the campaigns.
How are campaigns structured?
Tom: It’s actually pretty familiar if you’ve used Google or Meta before.
You’ve got your main campaign, then ad groups, and then the ads themselves. Nothing too complicated there!
How does targeting work?
Tom: Targeting is more context-based than keyword-based.
For example, if someone asks ChatGPT for the best things to do in Christchurch, tourism businesses could potentially show ads against that type of query.
We’re still testing different approaches, but at the moment, we’re finding it’s strongest for awareness, consideration, and research-stage activity rather than straight conversion campaigns.
What does a ChatGPT Ad look like?
Tom: The ads themselves are pretty small.
You get a small image and a short amount of text. With a maximum character count of 100, you don’t have a lot of room to work with.
The specs for the main image are at least 256 × 256 px. Ideally 512 × 512 px or bigger so it stays nice and sharp. Logo/favicon: Around 128 × 128 px works well.
How much budget would you start with?
Tom: It really depends on your budget and goals, but to start, at least $25 per day, which is also the platform minimum.
That gives the campaigns enough budget to actually do something meaningful and gather some data.
What conversion tracking needs to be set up?
Tom: Conversion tracking can be set up through Google Tag Manager.
It’s fairly similar in principle to Meta Pixel tracking. You can optimise towards conversions, clicks, or views, although we’d generally recommend focusing on conversions where possible.
Do I still need UTM parameters?
Tom: Yep, definitely.
We’re seeing some discrepancies between what ChatGPT reports and what comes through into GA4, so UTMs are still really important. I’d treat that as best practice for now.
How will traffic appear in GA4?
Tom: Generally, we can configure it to come through as ChatGPT Paid Traffic and tie campaign names through as well.
That said, we are seeing some drop-off between the traffic ChatGPT reports and what appears in GA4, so it’s something we’re keeping an eye on.
What mistakes will advertisers make when they first start using ChatGPT Ads?
Tom: Probably expecting too many conversions too quickly.
The traffic is really relevant and high quality, but I think a lot of the value sits around awareness, consideration, and being present when people are doing research. You might not be able to directly attribute every conversion back to the ads, so it’s important to be realistic with expectations.