ChatGPT Ads are Officially Now a Thing, and For Advertisers, They’re Cheaper Than Ever. Here’s What You Need to Know.

Since ChatGPT ads were introduced about a year ago, the costs associated with it limited the pool of advertisers to the select few who could cut a $200,000 check just to get in the door. As of this month, you can test into this new ad tech for just $25/day. That’s right, for less than $1000/month you can target users in an entirely new way.

Whether you’re looking to regain the kind of traction that you were accustomed to with traditional search, or simply looking to diversify into a new channel, it’s worth building an understanding of the targeting model before you get started. That gets you a long way towards deciding whether running ChatGPT ads is worth it for your business.

Ad Placement 

It’s important to understand that paying for a ChatGPT ad does not place your business inside ChatGPT’s answer or influence what the platform recommends. Instead, ads appear after the response as separate sponsored cards. They’re clearly labeled, visually distinct from the organic answer and typically include an advertiser name, headline, brief description, image and link to a landing page. Depending on the conversation, users may see one or more relevant ads, but the sponsored content remains separate from the answer itself.

Audience Targeting Options

First, it’s worth considering who is even going to see ads in the ChatGPT interface. Ads only show up for adult users on the free tier and low-cost Go tier. Anyone paying for Plus or higher won’t see ads at all. If your buyer profile is likely to have already upgraded, common amongst the most engaged, professional and higher income demographics, this channel is fundamentally misaligned with your marketing goals.

Additionally, if you’re an international brand, ads are currently only available in the US, UK, Canada, Australia, Japan, New Zealand and South Korea. And category-wise, financial services, gambling, alcohol and adult content are excluded. The same theoretically goes for legal services, but users have been reporting seeing ads from lawyers. It’s likely that, regardless of the industry you’re in, it’s worth trying to run ads to see whether the exclusions apply, and if so, what the nuances are.

Context, Not Keywords

It goes without saying that this ad technology is incredibly different from traditional search. OpenAI’s engine isn’t assessing the relevance of a record of the content from a page on your website against the kind of caveman non sequitur that is the keyword as we know it. In the past two decades we all learned to sort of grunt what we want into Google: “best running shoes,” “tapas restaurants in Phoenix.”

ChatGPT relies instead on a very open-ended set of parameters. You can use lists of keywords or topics, but it appears that OpenAI is weighting targeting recommendations toward what they call a “context hint.” This is a plain language description of the kinds of chats your product aligns with. If you’re advertising running shoes, that might be someone asking about marathon training and recommendations. 

What is the same is what I’ll call the “window of intent.” If you go super broad with your description, “people who are asking about comfortable shoes,” you’ll be competing against other running shoe companies, sure, but also the makers of slides, sandals, Crocs, clogs and even casual dress shoes. But go too narrow, say by describing “people asking for ultra-marathon trail running gear” and you’ll whittle your impressions down to almost nothing (because nationally, how many people actually need gear for off-road ultra-marathons? 175?).

The other thing that’s the same is the account structure. There, OpenAI leaned on the industry standard: campaign > ad group > ad. Structure is actually one the few levers you have to use to increase the relevance of your targeting. This is where you can play around with single-product listings vs. groupings and, of course, brand level messaging.

Relevance and What the ChatGPT Auction Rewards

There’s an auction. Of course there’s an auction. Which means that, at least in theory, part of what determines whether your ad shows or your competitors does, is how much you’re willing to pay. Even in the days of more transparency within both Google and Meta Ads, this was a dynamic that didn’t exactly favor advertisers. After all, the floor for costs was the floor, and how much exposure you got was based on your threshold for pain relative to what data you could get your hands on (which wasn’t much).

Just like its forebears in digital advertising, ChatGPT is asserting that it’s not just about whoever bids the highest. Instead, this is a “relevance-weighted auction,” which means that at least in theory it’s about relevance of context hint to chat x bid x landing page alignment.

In ChatGPT ads you have visibility into impressions, clicks, spend and conversion if tracking is set up. You’re going to want to set up tracking. Please, set up tracking. Without it, you’ll be launching dollars out into the ether. That’s a problem from both an immediate results perspective and from an audience refinement one. The AI is doing the matching, and that’s not going to get better if you’re not feeding it data about who is and isn’t likely to take an action.

How Much Do ChatGPT Ads Cost?

Right now, clicks are running $2-$5. If you bid under $3, the platform straight up tells you your ads won’t deliver. So there’s your floor. And all that data infrastructure sure does cost trillions to build, so I would expect those costs to do nothing but increase over time.

2026 Cross Industry Benchmarks*

*Prices fluctuate significantly across platforms and industries — actual costs depend on targeting, seasonality, competition, and creative quality. Treat these as directional benchmarks, not fixed rates.

Results and Reporting

You can run ads on ChatGPT. You can expect it to cost a certain amount for the time being. And you can do a lot of testing and experimenting to see what kind of prompts or lists or context improves performance. But when it comes to measuring that performance, your options are limited.

It’s currently possible to get numbers of impressions and clicks, plus spend (of course). You can also track conversions if you set that up. What you don’t get are demographic breakdowns, behavioral or affinity clues or insights into the types of conversions your ads are running in. 

Geographic insights are similarly opaque. Targeting is available down to the city level, but reporting is only available by country. That means that standard dashboards will not give you any real understanding of performance by market. You’re left to bake that into your campaign structure, and back into numbers by geography.

None of that is great. Quite frankly, there’s no real reason for this level of obfuscation. But the masterminds at OpenAI likely understand that our collective memories are short when it comes to Internet technology. They’re training us to lower our expectations now, so that they can continue to develop their ad tech how they see fit (and whatever nurtures their bottom line).

But if your target audience aligns with the free tier of ChatGPT users, there’s now no good reason to sit on the sidelines. Build an understanding of what’s effective for your audience, and build an intentional cost/benefit analysis of this new channel vs. what you’re currently running elsewhere.

If you need help, Sonar is here. We’ve been running our own ChatGPT ads for several months and have learned a lot about how this ad tech works, what targeting gets results, and the importance of conversion tracking. Seriously. See how we work alongside marketing teams to bring additional expertise to emerging channels like this one.