Digiday WorkLife Awards finalistAlwaysOn Digital is a finalist for the 2026 Digiday WorkLife Awards in Most Committed to Work/Life BalanceDigiday WorkLife Awards finalistSee the finalistsSee more →

AlwaysOn Digital

Should You Be Testing ChatGPT Ads in Q1? Six Questions to Ask Your Partner First

a cautious angler in a small rowboat dips a single short fishing line into a glowing new pond, while on the shore an overexcited man tips an entire wheelbarrow of gold coins into the same water.

Advertisers in the US spent an estimated $70.7 million on ChatGPT ads in October, more than double April’s figure, according to AdClarity. So if you haven’t been asked about testing ChatGPT ads yet, you will be in your Q1 planning meetings. Our view is simple: for the right brands, a small, deliberate test in Q1 makes sense. Moving real budget out of channels that already work does not. And most of the tests we expect to see this winter won’t prove anything either way.

ChatGPT ads are real money now

This isn’t a closed pilot anymore. OpenAI dropped its spending minimum when it opened a self-serve Ads Manager to US businesses in May, after launching at a six-figure minimum and a $60 CPM. Advertisers can now buy on cost per click, with a suggested starting bid of $3 to $5. Ads sit below the answer, are labeled, and show only to users on the free and lower-cost plans.

That low barrier is exactly why testing ChatGPT ads needs a grown-up conversation. When anyone can switch a channel on with a credit card, the question stops being “can we get in?” and becomes “will we learn anything worth the money?”

Why most tests will prove nothing

Three things make ChatGPT different from the search and social buys you already know.

First, targeting is contextual. Advertisers give the system context hints and a landing page, not keywords, and there’s no retargeting. You can’t steer it the way you steer a search campaign.

Second, the platform holds the wheel. OpenAI makes the delivery decisions, and as of its May launch it had promised third-party measurement without naming partners or a timeline.

Third, the numbers don’t always line up. One agency that published its own test reported 57 clicks in OpenAI’s dashboard and fewer than 20 visits in Google Analytics. One campaign isn’t a verdict, but it’s a warning. If nobody decides before launch which numbers count, the test ends in an argument instead of an answer.

Questions to ask before testing ChatGPT ads

Take these to your in-house team or your media partner. The answers tell you more than any pitch deck.

  1. Why is our brand a fit for this, specifically? A good answer ties it to how your customers research: people who ask detailed questions before they buy. A vague answer is “it’s the hot new channel.”
  2. What’s the one question this test answers? Good: “Does ChatGPT bring us new customers at a cost per lead within 20% of paid search?” Vague: “We’ll see how it performs.”
  3. Which number is the source of truth? Good: they’ll name it before launch and explain how they’ll reconcile OpenAI’s reporting with your analytics and your CRM. Vague: “We’ll report what the platform shows.”
  4. Where is the money coming from? Good: a defined test budget that doesn’t starve a proven channel, or a clear holdout so you can see what shifted. Vague: “We’ll pull a little from search.”
  5. Whose account is it? OpenAI expects a separate account per advertiser. Good: the account is in your name and you have access. Vague: “We run it through ours.”
  6. When do we decide? Good: a fixed end date and a go, no-go or extend call written down in advance. Vague: “Let’s give it a few months.”

Questions to ask yourself

Before you judge your partner, check your own side. Could you say tonight, without asking anyone, what a lead from a new channel is worth to you? If not, no ChatGPT test will satisfy you, because there’s nothing to compare it to.

Then look at your partner’s record. Did they bring up ChatGPT ads before you asked, with a point of view on whether it fits your business? Or did they wait, and then say yes to everything? A partner who only reacts to your questions is a partner who isn’t thinking about your account between calls.

And ask who’s actually in your accounts. A new channel that a junior coordinator sets up in an afternoon is a new channel nobody senior is watching.

What a good test looks like to us

We think new inventory deserves curiosity and discipline in equal measure. A good plan for testing ChatGPT ads has a written hypothesis, a budget you can afford to lose, a measurement plan agreed before the first dollar goes out, and one senior person who owns the call at the end. It runs alongside your proven channels, not instead of them.

That’s the standard we hold ourselves to. It’s the same reason we’ve argued that your campaign isn’t “AI-powered,” it’s human-guided. The platforms keep adding automation, such as OpenAI’s conversion-optimized bidding added in July. Someone with judgment still has to decide whether the results are real.

Compare it with your current setup

If your current partner gave you clear answers to the questions above, good. Keep them. If the answers were vague, or the conversation never happened, let’s compare notes. We’re glad to scope a Q1 plan for testing ChatGPT ads alongside your current setup, run a side-by-side review of how your search and social are measured today, or just walk through your plan. Talk to our team when you’re ready.

Put 19 years of judgment on your campaigns.

Tell us the KPI you’re measured on. We’ll show you how our team would hit it.