Running ChatGPT Ads for clicks: what CPC is good for and when to leave it
why most advertisers should start on CPC, what makes creative work next to an answer, and the signal that tells you it is time to move to conversions.
the short version: CPC is where you should start, not because clicks are the goal but because it is the cheapest way to find out what people respond to before you ask an algorithm to optimise anything.
Why start here
the conversion objective needs conversion volume to learn from. on day one you have none. launching straight into oCPC means the system is optimising toward a handful of events, which is close to optimising toward noise.
CPC gets you click data fast, and click data tells you the thing you most need to know early: which framing of your product makes a person stop and act. you can be learning that while your conversion tracking accumulates enough events to be worth optimising against.
Creative next to an answer is a different job
this is the part that surprised me most coming from paid social.
in a feed, the user is browsing. your job is to interrupt. hooks work, pattern breaks work, curiosity works.
in ChatGPT, the user is mid-task and slightly impatient. they asked a question and they are reading the answer. an ad that tries to interrupt that reads as an obstacle. what works instead is an ad that looks like the obvious next step for someone who just read that answer.
concretely: lead with what the thing does, not with a clever line. name the specific problem in the same language the person just used. skip the hype, because it sits next to a factual answer and the contrast makes it look worse.
the ads of mine that worked best were the ones that would have been boring in a Meta feed.
Watch cost per click, but do not optimise for it
a falling cost per click feels like progress and often is not. you can always get cheaper clicks by reaching people less likely to care.
the number that matters is what happens after the click. put your CPM, click rate and post-click conversion rate into the ad funnel calculator and it will tell you which of the three steps is actually costing you. a low click rate is a creative problem. clicks that do not convert is a landing page problem. these need completely different fixes, and the most common mistake is making new ads when the landing page was the issue.
How much to test
enough variants that a result means something, and few enough that each gets real volume.
the arithmetic is unforgiving: split a small budget across too many ads and every result is inside the range chance produces, so you pick a winner at random and learn nothing. the creative testing planner tells you the most you can properly test at your budget, which is usually fewer than people expect.
for a first ChatGPT Ads campaign I would run a handful of genuinely different angles rather than many variations of one, because at this stage you are looking for the right message, not polishing a message you have already found.
When to move to conversions
move when you have enough conversion events flowing that the system has something real to optimise toward, and when you know which angle works well enough that you are no longer searching.
remember you cannot convert an existing campaign. you build a new one with the objective set to conversions, so treat the CPC phase as deliberately temporary. the optimisation side is in running ChatGPT Ads for conversions.
The wider point
ChatGPT Ads is cheap right now because it is new and most advertisers have not arrived. that is a window, not a moat, and every channel decays as it fills up.
so run it hard while it is cheap, and do not build a company that depends on one channel staying cheap. the paid ads playbook covers the testing discipline in more depth, and the full playbook library has the other fifteen channels for when this one normalises.