We run live campaigns on ChatGPT for our clients. Ads on ChatGPT can bring real leads, with traffic quality that holds up well against the channels you already use. They also come with their own rules, and those rules decide a lot before your first ad ever runs.
Here is what we have learned, in plain terms, for any advertiser thinking about testing the channel.
Check the ad policy before you build anything
It is a good place to start.
OpenAI assigns every ad to a policy category. Some categories are restricted in some countries. If your product falls into one of them, your ads may simply never run, even when the campaign looks perfectly set up.
Not every category is allowed in every country. So before you plan ad groups, write copy or design creative, read the ad policy for your category and your target market. It takes little time and can save you a full campaign build that never gets a single impression.
If your category is restricted, treat that as a starting point for the next step below.
Lead with value and let the product follow
We saw a clear pattern in what gets approved. Ads that use direct product terms can get blocked in a restricted category. Ads built around a gift or a clear piece of value can pass the review and run.
That kind of message also tends to work well on its own merits. On ChatGPT, people are in the middle of a conversation, looking for help. An ad that offers something useful fits that moment. An ad that sounds like a sales pitch for a regulated product stands out in the wrong way.
A few practical ideas:
- Offer a guide, a course, a checklist or a tool that relates to your product.
- Keep the message clear, friendly and concrete.
- Let the landing page do the work of explaining the product.
Always stay within the policy. The goal is to communicate value, and the review still applies to everything you publish.
A simpler platform built around conversations
Setting up a campaign on ChatGPT is simpler than on Google or Meta. There are far fewer settings to manage. The structure will feel familiar: campaigns, ad groups and ads.
The ad itself is a card that appears inside the chat, with an image, a short headline, short body text and a link.
The biggest change is targeting. Each ad group is described by its audience and by the conversation topics you want to appear next to. You think about what people are talking about and asking for help with, and you describe that. Keyword lists are not part of the setup.
In practice, this means your planning starts from the questions and situations your customers bring to ChatGPT. Build ad groups around those themes, keep each one focused, and write ads that match the topic.
Traffic quality and cost sit in the middle
When we compare ChatGPT traffic with other channels in Google Analytics, it lands in a sensible place. Visitors engage more than typical Meta traffic and less than high intent Google search traffic.
Costs follow a similar pattern. ChatGPT sits between Meta and Google search: above Meta and below the most competitive search terms.
That makes it a reasonable channel to test alongside your core accounts. It will not replace strong search campaigns, and it gives you access to people at a moment when they are actively looking for answers.
Separate click conversions from view-through conversions
We install the OpenAI pixel through Google Tag Manager and track the events that matter, such as leads.
Here is the catch. The platform reports conversions that came from a click, and it also counts view-through conversions, where someone saw the ad and converted later without clicking. Together, these can make results look much stronger than what you see in your own analytics.
Read click conversions on their own. In our experience, those are the numbers that line up with Google Analytics, and they are the ones to use for budget decisions. Treat view-through as a secondary signal.
Limited visibility into where ads appear
You cannot see which prompts your ads showed up on. Reporting is by ad group, so you learn how each topic group performs as a whole.
This affects how you set up and read a test:
- Keep ad groups narrow so each one tells you something useful.
- Change one thing at a time.
- Give the test enough time before drawing conclusions.
Who should test ChatGPT ads now
A test makes sense when your core channels are already tracked and structured well, when there is room for a separate experiment budget and when someone owns the test and reviews the results regularly.
If you are not sure your accounts are in shape, our post on what to check in a Google and LinkedIn Ads audit is a good place to start. We look at the full paid mix together, which we covered in what a full stack paid media team does.
Our take
Advertising on ChatGPT works. It brings leads, the traffic quality is solid and the setup is easy to manage. To get the most from it, check your category policy first, build your message around value, read click conversions separately and plan your tests around topic level reporting.
If you want to know whether ChatGPT, or another channel, makes sense for your accounts, we can look at them together.