ChatGPT Ads: What We Learned Spending $1,524 on OpenAI’s Ad Platform

ChatGPT ads are here. And with them comes a whole new set of questions for marketers.

How much do they cost? How does targeting work? Are people actually clicking on them?

And perhaps most importantly, can they generate results comparable to what we’re used to seeing from Google Ads, LinkedIn and other established advertising platforms?

Between June 17 and August 28, 2026, Brafton’s paid media team spent just over $1,500 running a campaign through OpenAI’s advertising platform.

Our Director of Paid Media, Zach Shah, documented the experience, from campaign setup and performance to targeting and reporting limitations.

So, what did $1,500 actually get us? And what can other marketers learn from our experience?

Our ChatGPT Advertising Experiment: The Setup

Zach kept things relatively simple with the setup.

Goal: To drive traffic to Brafton’s website from US-based users who might be interested in our marketing services.

We weren’t trying to test every possible campaign configuration or build an elaborate advertising funnel. We wanted to get a feel for the platform, see what kind of results it could generate and establish a baseline for future testing.

Here’s how we configured the campaign:

In hindsight, this relatively simple setup was also one of the experiment’s limitations. With just one ad group and one creative, we had limited opportunities to test different messages against different conversational contexts. A more varied campaign structure could help us better understand where ChatGPT advertising works best.

Something to note: Unlike Google Ads, where you can build campaigns around specific keywords and search terms, ChatGPT uses something called context hints.

Essentially, you describe the conversations, topics and customer needs where your advertisement might be relevant. OpenAI then uses those descriptions to help determine when to serve your ad.

Our context hints covered several Brafton services, including B2B content marketing, SEO, AI search visibility, blog writing, paid search management and demand generation.

More on context hints later.

The Results: What $1,524 Got Us

After 72 days, here’s what our campaign generated:

Not exactly groundbreaking results, but not entirely without promise, either.

We generated more than 46,000 impressions and 263 website visits, so there was clearly an audience available to reach.

Our average cost per click was $5.80, which is comparable to certain established advertising channels.

What’s less exciting is that of those 263 clicks, only one resulted in an attributed conversion. That puts our cost per lead at $1,524.11.

Now, one campaign with a relatively small budget isn’t enough to draw sweeping conclusions about an entire advertising platform. There are plenty of variables that can influence performance, including campaign objectives, creative, targeting and landing pages.

Still, the conversion numbers weren’t the only thing that caught our attention. Some of the most interesting findings came from what we couldn’t see or control, which led to 5 lessons you might like to know.

Lesson 1: ChatGPT’s Reporting Leaves Lots To Be Desired

Anyone who’s spent time managing Google Ads knows just how much information it makes available: search terms, impression share, auction insights, device performance, conversion data and countless other metrics. You might not need all of them for every campaign, but they’re there when you do.

In comparison, OpenAI’s Ads Manager, at the time of our experiment, offered just six core delivery metrics:

• Impressions.
• Clicks.
• Spend.
• CTR.
• CPC.
• CPM.

Conversion reporting was available separately, and advertisers could break down performance by device, country and platform.

But there was no search term reporting, no auction insights, no impression share data and no detailed breakdown of the conversational topics driving impressions.

Why is that a problem? Well, imagine you’re running a Google Ads campaign and want to know why you’re not getting more clicks.

Are your bids too low? Are competitors outranking you? Are your ads showing for irrelevant searches? Are you missing out on impressions because of budget limitations?

Those are all questions you can investigate using Google’s reporting tools. With ChatGPT, we couldn’t confidently answer them.

It’s worth noting that OpenAI has since expanded its measurement capabilities. In October 2026 — a couple of months after our test concluded — the company announced new integrations with attribution, analytics and incrementality measurement partners to help advertisers better understand campaign performance. These improvements don’t necessarily address every reporting limitation we encountered, but they do represent progress toward more comprehensive measurement.

Some of the Numbers Didn’t Add Up, Either

Beyond the limited reporting options, Zach also noticed some unaccounted-for impressions.

Our campaign generated 46,024 impressions. But when we looked at the platform breakdown, the individual rows accounted for only 21,509. That left 24,515 impressions — approximately 53% of the total — unaccounted for in that particular report.

The spending figures showed a similar discrepancy, with roughly 56% of our total spend missing from the platform-level breakdown.

To be clear, this doesn’t necessarily mean impressions or advertising dollars disappeared. It means the numbers presented in different parts of the interface didn’t reconcile.

“I don’t think this is deliberate obfuscation,” explains Zach. “It reads like a beta reporting pipeline where segment rows are computed from a partial event stream. But the effect on an advertiser is identical: you cannot audit your own spend.”

If you’re responsible for managing advertising budgets, you need to be able to explain where that money went and what it accomplished. Without reliable reporting, that’s much harder.

Lesson 2: Context Hints Aren’t the Same as Keyword Targeting

In traditional paid search, advertisers can target specific keywords, review the search terms that triggered their ads and exclude irrelevant queries.

But with ChatGPT Ads, advertisers can only provide a written description of the conversations where their ads might be relevant. That’s what they’re calling context hints. And in practice, they’re essentially suggestions and not strict targeting instructions.

Fair enough: OpenAI’s interface explicitly describes them as guidance for matching rather than exact-match rules.

So, while we could tell the platform what kinds of conversations interested us, we couldn’t guarantee our ads would appear only in those contexts. Nor could we see which specific topics generated our impressions.

That creates problems for optimization. Because let’s say our ads were performing well in conversations about AI search visibility but poorly in conversations about general content writing.

In Google Ads, we’d have ways to identify that difference and adjust our targeting accordingly. ChatGPT doesn’t offer that same level of visibility into the conversations triggering ads, although advertisers can compare performance across ad groups and creative to better understand which messaging and customer needs generate results.

Obviously, there’s a reasonable argument for keeping conversational data private. Users shouldn’t have to worry about advertisers accessing their personal chats.

But, as Zach explains, “The practical consequence is that the single most valuable optimization loop in paid search — read the search terms report, add negatives, split out winners — does not exist here, and there is no announced plan for a privacy-safe substitute.”

OpenAI has said only that it may “explore additional measurement insights over time while continuing to protect user privacy.”

Lesson 3: ChatGPT Ads Are Priced Like Search but Behave More Like Paid Social

When Zach compared our results with industry benchmarks for Google Search and LinkedIn Sponsored Content, he noticed that our average CPC was almost identical to the Google Search benchmark for business services, while click-through rate was considerably lower.

In fact, our 0.57% CTR fell right within the range typically associated with LinkedIn Sponsored Content. And our conversion rate was substantially below the Google Search benchmark:

Note: Google and LinkedIn figures are third-party industry benchmarks, not matched campaigns. These comparisons are directional rather than exact.

“ChatGPT ads are priced like search, but seem to behave like paid social,” Zach said.

Let’s think about that: When someone searches Google for “SEO agency near me” or “B2B content marketing services,” for example, there’s a reasonable chance they’re actively looking for a provider. That search carries some degree of commercial intent.

But someone having a conversation with ChatGPT about improving their website traffic might simply be looking for information. They could be researching a problem, brainstorming ideas or learning about a subject without any immediate intention to purchase anything.

An advertisement appearing alongside that conversation might be relevant, but relevance and purchase intent aren’t the same thing.

“In paid search, you pay a search CPC because the click carries search intent. Here we paid a search CPC for a click that behaves like a feed click,” Zach explained.

That’s not to say ChatGPT users never have commercial intent. But without more detailed reporting, advertisers have limited insight into which conversations are driving clicks and whether those users are actually in the market for their products or services.

At least for now, it may make more sense to evaluate ChatGPT advertising through a paid social lens rather than treating it as a replacement for search advertising.

Lesson 4: Most of Our Traffic Came From Mobile Devices

The overwhelming majority of our campaign activity happened on mobile:

• 82% of impressions came from mobile devices.
• 89% of clicks came from mobile.
• 87% of total spend went toward mobile traffic.

Mobile also generated a higher CTR than desktop: 0.62% compared with 0.36%.

Device usage can tell us something about how people are engaging with advertising. Someone browsing ChatGPT on their phone might be researching a topic during their commute or casually exploring an idea. Someone on a desktop might be more likely to be working through a specific business problem.

Of course, those are just possibilities and not conclusions we can draw from device data alone — which is kind of part of the problem here. We could see the difference in performance, but we couldn’t adjust bids by device to test whether shifting more budget toward desktop users would improve results.

That’s another example of how ChatGPT’s reporting and optimization capabilities were somewhat disconnected during our experiment.

So, Should Marketers Invest in ChatGPT Ads?

If you’re already generating strong, measurable results from Google Ads, LinkedIn or another established platform, Zach doesn’t recommend moving a significant portion of that budget to ChatGPT just yet.

But not because there isn’t opportunity; our campaign generated more than 46,000 impressions on a relatively modest budget. The bigger question is whether advertisers can reliably turn that inventory into business results and demonstrate the value of their investment.

“Don’t move budget out of Google or LinkedIn for this just yet,” Zach said. “Not because the audience isn’t there […] But because you cannot currently prove what you bought, and a channel you can’t audit is a channel you can’t defend in a QBR.”

That said, there’s certainly a case for experimentation.

If your business has a dedicated testing budget and you’re comfortable with some uncertainty, ChatGPT advertising could be worth exploring. Just don’t expect the same level of control, transparency or performance you’re accustomed to elsewhere.

Five Recommendations for Advertisers Considering ChatGPT

If you’re interested in testing ChatGPT ads, here are five recommendations based on this experience:

1. Set aside a dedicated testing budget: Avoid shifting spend away from proven campaigns. Give yourself enough budget to gather meaningful data, rather than drawing conclusions from a handful of clicks.
2. Diversify your ad groups and creative: Build around different customer needs and use cases, with multiple creative variations to test what resonates.
3. Create dedicated landing pages: Direct users to pages that address the specific problems or questions that might have prompted them to click your ad.
4. Look beyond last-click conversions: Consider assisted conversions, branded search activity and website engagement to better understand the channel’s potential value.
5. Make sure your tracking works: Verify conversion events, UTM parameters and analytics before launching. Consider additional measurement tools, such as OpenAI’s Conversions API.

What’s Next for ChatGPT Advertising?

Our first ChatGPT Ads raised some interesting questions about where paid media might be heading.

For years, search advertising has revolved around keywords, bids, audience signals and increasingly sophisticated automation.

ChatGPT introduces something different: the ability to describe the kinds of conversations where your business might be relevant. And it’s already evolving, too.

ChatGPT Ads now supports conversion-optimized campaigns, allowing advertisers to optimize toward specific actions rather than just clicks. OpenAI has also announced new visual ad formats and expanded measurement partnerships, giving marketers more ways to reach potential customers and evaluate results.

“This is the first ad platform where the targeting input is prose rather than keywords, Zach said. “That is a content and messaging discipline, not a bid-management discipline.”

If conversational context becomes an increasingly important signal for ad targeting, marketers may need to spend more time thinking about how they describe their customers’ problems, needs and motivations.

In other words, the quality of your messaging could become just as important to targeting as the technical configuration of your campaigns.

Does that mean traditional PPC expertise is going away? Not at all. But it could mean that the skills involved in building effective advertising campaigns will continue to evolve as AI platforms develop.