ChatGPT Ads Require a New Targeting Mindset

It was about control.

Paid media managers are trained to add precision: more targeting, more exclusions, more audience detail, more campaign structure. But our tests –one week, a $100 daily budget, and limited data–taught me almost the opposite.

In a conversational ad environment like a chatbot, too much targeting detail may not make the campaign smarter. It may just give the model less room to deliver.

It’s an early signal about how differently this channel may need to be managed.

TMI isn’t good for ChatGPT

The most interesting part of ChatGPT Ads is the context-hints section.

Instead of choosing traditional keywords or building a detailed audience, advertisers describe the types of conversations where their product or service might be relevant.

Naturally, I overdid it.

I included age, gender, industries, business types, pain points, use cases, and a long list of questions someone might ask ChatGPT. What was supposed to be a simple context hint quickly turned into a dense block of targeting instructions.

In theory, I was giving the model everything it needed.

In practice, I may have given it nowhere to go.

The campaign barely delivered and struggled to spend even a small portion of the daily budget.

So I simplified the setup.

I removed most demographic restrictions, cut the long list of possible questions, and focused on four things: business type, user intent, pain points, and practical use cases.

The revised version was much shorter.Delivery improved almost immediately, and the campaign began spending consistently.

This was not a controlled experiment, so I cannot claim that shorter context hints will always perform better. Other platform factors may have changed at the same time.

But the directional lesson matters:More detail does not always create more precision. Sometimes it just creates more ways to limit delivery.

Paid media managers are trained to control targeting. We build audiences, exclusions, keyword lists, and campaign structures designed to reduce uncertainty.

ChatGPT Ads may require a different mindset. Instead of telling the platform exactly who the user is, we may need to explain what problem the user is trying to solve.

That is a small change in setup, but a meaningful shift in strategy.

Human Faces > Logos

The second learning came from the creative.

The initial ads used company logos, service graphics, and traditional branded assets. The click-through rate was around 0.5%. Then I replaced those assets with professional headshots and personal-brand-style photographs.

CTR increased to approximately 5%. That is a big difference.

Again, this is early data. The investment is limited, the sample size is still small, and the test was not perfectly controlled. But the gap is large enough to pay attention to. 

My working theory is that human faces feel more natural inside a conversational environment. When people use ChatGPT, they are usually not looking for ads. They are asking a question, researching a problem, or trying to make a decision. A logo can feel like a brand entering a conversation where it isn’t particularly welcome.

A human face can feel like a person joining it.That difference may influence attention and trust.

This does not mean every advertiser should immediately replace all branded creative with headshots. The strongest approach will depend on the industry, the offer, and the person featured in the image.

But it does suggest that creative built for Google Display, Meta, or LinkedIn may not automatically translate to ChatGPT.

The environment matters. The user mindset matters. And the creative should feel like it belongs in that experience.

The Most Important Result Is Still Missing

A higher CTR is encouraging. But clicks are not the goal.

The real question is whether those clicks lead to qualified traffic, meaningful engagement, and eventual conversions.

Right now, I do not have enough data to answer that. That is why I am not calling the campaign successful yet.

The next phase of the test will focus less on whether the platform can spend and more on the quality of the traffic it produces.

I want to understand which conversation themes drive stronger engagement, whether certain context hints attract higher-intent users, and whether ChatGPT Ads can influence customers earlier in their decision journey.

For now, the early takeaway is simple: Do not treat ChatGPT Ads like a smaller version of Google Ads.

Give the model enough direction to understand relevance, but not so many restrictions that it has nowhere to go. And in a conversational environment, do not underestimate the power of showing a real person.

One week is nowhere near enough time to judge how the channel truly performs. At this stage, I have early signals, not a verdict.

But those signals point in the same direction: ChatGPT Ads may not reward the paid-media instinct to over-engineer the audience. In a conversational environment, the better starting point may be the problem someone is trying to solve.

The next question is whether that shift produces better traffic, stronger engagement, and eventual conversions. That is what I’ll be watching as the campaign collects enough data to tell a clearer story.