By Monica Shukla, Mile Marker, Vice President of Biddable COE | Originally Published on MediaPost on September 24th, 2026.

The most important number in ChatGPT advertising right now isn’t CPM, CPC or click-through rate. It’s the absence of a reliable benchmark.

OpenAI itself says ChatGPT Ads is still in beta, with no performance benchmarks yet across advertisers, industries or campaign types. At the same time, the platform is expanding fast: self-serve Ads Manager, CPC buying, conversion measurement and increasingly conversion-optimized campaigns. That combination — rapid development, limited evidence — should give marketers pause about the channel’s own reporting. 

What Signals to Listen to

We’ve run structured pilots, and our clearest finding is this: a ChatGPT campaign can look weak in platform reporting while generating signals of genuine demand elsewhere. That isn’t an argument to ignore performance data; it’s an argument to understand what each source can and cannot tell you.

ChatGPT isn’t simply another search or social environment. The advertising opportunity exists while people explore options and make decisions inside a conversation, so the journey may not end with the click. A user can see an ad while researching, leave ChatGPT, search the brand, and convert days later elsewhere. A framework built around last-click conversion can miss that journey. 

What Last Click Models Miss

The first discipline, then, is defining the channel’s job before its KPI. If the strategic role is consideration and demand creation, a week-one ROAS target is the wrong test. Before launch, marketers should agree whether the objective is qualified attention, incremental demand, direct conversion, or some blend. Our pilots suggest a four- to six-week learning period beats an absolute early target: look for improving engagement, better delivery quality and declining unit costs, since conversions may lag.

The second discipline is triangulation. Platform reporting matters but shouldn’t be the entire measurement system. Ads Manager reports impressions, clicks, spend, CTR, CPC, CPM and conversions; the Pixel, Conversions API and UTMs connect activity to downstream behavior. Reconciliation — not another dashboard — is the core capability. In our testing, the platform’s strongest-looking week was sometimes among the weakest for genuine new demand, and vice versa: one source alone gives an incomplete picture. 

Three Checks You Can Control in Your Test

Three checks help. First, qualify the traffic: a click isn’t a valuable visit, so compare it against first-party analytics. Second, watch branded search — a rise in brand queries during a test can reveal a consideration effect campaign reporting misses. Third, build an experimental baseline: at modest budgets, incremental demand can be too small to separate from noise, so run a controlled on-off spend test rather than manufacturing a lift or writing off the channel.

Creative matters too: test product-feed and manually built creative, since the winning format in search or social won’t automatically win here. OpenAI’s guidance favors conversational relevance, clear benefits and multiple variations over keyword matching. And set baseline measurement before the first dollar is spent — once a test begins, that data can’t be recreated.

The emerging opportunity in conversational advertising isn’t just a new audience; it’s a new point in the decision journey. Advertisers who win won’t necessarily have the biggest budgets or lowest CPCs — they’ll be the ones who can separate platform performance from business impact and reconcile the two.

ChatGPT advertising is getting easier to buy, and measurement is improving, but there’s still no universal scorecard. That uncertainty isn’t a reason to sit out — it’s incentive to enter with a better measurement system than everyone else. 

Read the original post on MediaPost 

Ready to navigate the evolving media landscape? Let’s talk.