ChatGPT ads look like a test channel, not a settled search replacement

TL;DR: Early ChatGPT ad data points to test budgets, not a channel verdict: CTR crossed 1% in four markets but not the U.S., and advertiser churn means the format is still finding its buyers.

Are ChatGPT ads working yet?

The primary source here is Matt G. Southern’s Search Engine Journal report, “ChatGPT Ad CTR Topped 1% In 4 Markets, Not In The U.S.: Report,” which cites Similarweb panel data on ChatGPT ad click-through rates by market and tracks how often the top advertisers changed.

The useful signal is not “ChatGPT ads work.” Too broad. The useful signal is narrower: in some markets, users are clicking at rates that look real enough to study. Search Engine Journal reported that ChatGPT ad CTR topped 1% in four markets, but not in the U.S.

That matters because the U.S. is usually where marketers look first for proof. If the U.S. underperforms while other markets clear 1%, this is not a clean global adoption story. It is a market-by-market behavior story.

And that fits the product surface. A chatbot is not a search results page. The user is often mid-task, not scanning links. Intent can be deeper, but less standardized. A user asking for help planning a purchase, comparing tools, or solving a workflow problem may be receptive to a relevant commercial option. A user asking for code, medical context, or a writing critique probably is not.

So CTR alone does not settle much. It tells us the ad unit can get attention. It does not tell us whether the click was useful, whether the advertiser got a qualified visitor, or whether the placement improved or degraded the assistant experience.

Why does advertiser churn matter?

The more interesting part is churn among top advertisers. Search Engine Journal reported that the top ChatGPT advertisers changed often.

That is usually a sign of experimentation. Brands are testing creative, audiences, budgets, and category fit. Some may be seeing enough to keep poking at it. Others may be cycling out because the unit is not yet predictable.

For operators, that is the real story. A new ad channel becomes important when three things line up: repeatable intent, measurable attribution, and enough inventory to matter. ChatGPT-style placements may have the first one in pockets. The second is likely messy. The third depends on user behavior and platform policy.

This is also where marketers need to resist lazy comparisons to Google search. Search ads became dominant because the query, ad, landing page, and conversion event formed a tight loop. AI answers blur that loop. The assistant may satisfy the user before any click. Or it may steer the user with a recommendation that feels more like advice than an ad.

That makes quality control more important. Bad ads in search are annoying. Bad ads inside an assistant can feel like polluted judgment.

What should builders and marketers do with this?

I would not treat this as a reason to move serious budget away from proven search, SEO, email, or affiliate channels. I would treat it as a reason to start instrumenting AI-origin traffic carefully.

I run Lucky Domains, whose work includes SEO and search visibility, and this is exactly the kind of channel shift that should be tracked before it becomes obvious in standard dashboards:

The practical move is simple. Segment traffic from AI surfaces where possible. Watch landing-page behavior, not just clicks. Build pages that answer comparison, buying, and implementation questions clearly, because assistant-driven users may arrive with more context than a normal search visitor. If you test ads, keep the test small and judge it by downstream quality, not novelty.

The catch most readers miss: a 1% CTR in an AI assistant is not automatically better or worse than a 1% CTR somewhere else. The surrounding task matters. A click from a user who just had a ten-turn buying conversation may be worth more than a generic search click. A click from a curious user poking at a new ad surface may be worth less. Treat ChatGPT ads as an intent lab first, and a scaled acquisition channel only after the conversion data earns that label.