Advertising on ChatGPT: What businesses need to know before getting started

These days, instead of typing "best urban bike Quebec" into Google, your potential customers are more likely to open ChatGPT and ask it to help them choose. In just a few seconds, the AI compares options and helps them make a decision. And in recent months, something new has been added to this experience: a brand can now appear below the response, in a clearly labelled sponsored content unit.

With over 900 million weekly active users, ChatGPT has become a true space for research, comparison, and decision-making. OpenAI has therefore gradually opened its platform to advertising since February, transforming the conversational tool into a new media channel for businesses in select countries, including Canada.

This growing new channel that is advertising on ChatGPT raises important questions for businesses: How does it work? Who is it for? How much does it cost? And most importantly, is it already worth integrating into your media strategy?

We break it all down for you, point by point.

What principles govern advertising within ChatGPT?

OpenAI launched its first advertising tests in early 2026 with the stated goal of developing this new revenue model without compromising the user experience.

Three principles guide this approach:

• AI-generated responses remain independent from advertisers
• User conversations remain confidential
• Ads are clearly identified as sponsored content

What does an ad in ChatGPT look like?

Unlike social media, where ads often seek to immediately grab attention, ChatGPT takes a much more understated approach. Ads never interrupt the conversation. They appear below it, clearly labelled as sponsored content, in a separate unit.

The format presented by OpenAI at the launch of the advertising platform is straightforward:

• An image
• A headline
• A short description
• a link to the advertiser's website

The goal is to offer an additional resource when it seems relevant, without diverting attention from the conversation.

For retailers, this basic format can also leverage a product catalogue. OpenAI then automatically generates ads from individual product listings and matches them to relevant shopping conversations—a mechanism similar to "dynamic product ads," akin to what Google Shopping or Meta already offer. This feature is also described by StackAdapt in its advertiser documentation. So it's not necessarily a single static thumbnail for every advertiser.

It's also worth emphasizing that ChatGPT's responses are not influenced by advertisers. OpenAI calls this principle response independence: the AI model first produces its response, and then a separate advertising system determines whether a relevant ad can be displayed. In other words, a business cannot "buy" an AI recommendation. It can only appear alongside a response, never within it.

What really changes: You're no longer targeting people—you're targeting intent

This is probably what most distinguishes advertising on ChatGPT: There is no audience targeting in the traditional sense.

If you're used to working with Google Ads, Meta, or LinkedIn Ads, you're used to selecting an audience. You define an age range, a location, interests—sometimes even a list of existing customers or visitors to retarget. In ChatGPT, this logic all but disappears.

Instead of selecting a traditional audience, advertisers describe the situations in which their offering is relevant. OpenAI calls these descriptions context hints: An approach that represents the main shift from traditional advertising platforms. As an advertiser, you describe in natural language the type of conversation in which your product or service would be relevant.

For example, you would no longer say: "I want to reach women aged 30 to 45 who are interested in cycling." Instead, you describe the situations in which your offering is relevant:

• Someone looking to choose a bike for their daily commute
• Someone comparing several models before buying
• A user looking to replace their car with a more eco-friendly mode of transportation

OpenAI's system then reads the actual content of the ongoing conversation (the intent, the progression of the dialogue, and the topic) and decides, in real time, whether your ad matches. It tries to understand the user's intent rather than find an exact keyword match. The distinction may seem subtle. In reality, it completely changes how a campaign is designed. The goal is no longer to find the right person, but the right moment in a conversation.

What parameters can you actually target in ChatGPT Ads?

Unlike traditional advertising platforms, the available parameters remain intentionally limited. Specifically, the only levers you control are:

• The geographic targeting, currently at the country level
• The context description, which replaces the keyword
• The budget;
• The bid type (cost per thousand impressions or cost per click, depending on the objective)

What you don't control (and will never see) are the internal signals OpenAI uses to refine relevance:

• The content of the active conversation thread
• The language
• The general location
• The conversation history for users who have opted into personalization

This is a deliberate choice by OpenAI to preserve trust: Conversations remain private, period. It is therefore intent- and context-based targeting rather than audience-based targeting—which requires a real shift in mindset for teams accustomed to Google or Meta.

How do you advertise on ChatGPT?

There are two ways to proceed:

• The first is self-service (Ads Manager Beta), managed directly by OpenAI, open to businesses of all sizes, but with a technical setup to handle on your own (context hints, bidding, pixel, business verification, etc.).

  
• The second is through technology partners like StackAdapt (a Tink partner), one of the first partners in OpenAI's advertising pilot program, which allows you to manage a ChatGPT campaign from a platform already used for other channels (connected TV, display, native, etc.).

Note that access to advertising on ChatGPT is not automatic for all industries. Categories deemed sensitive—such as health, politics, finance, or legal—are subject to stricter manual review, and some remain entirely excluded for the time being. It's therefore best to verify your industry's eligibility before building a strategy around this channel.

For many organizations, the main challenge is understanding how to build a first relevant campaign: Defining the right context hints, choosing a bidding strategy, measuring results, and making the necessary adjustments.

What budget should you plan for an initial test?

There is no longer a mandatory minimum to access the self-service platform; any eligible business can therefore start with a modest daily budget. That said, to gather enough data to determine whether this channel works for you, it's recommended to plan for the equivalent of a few thousand to a few tens of thousands of dollars, spread over two to three months, rather than running a single token test of a few hundred dollars that won't generate enough volume to draw a meaningful conclusion.

Is it expensive compared to other channels?

When it launched in February 2026, the channel was reserved for a small group of advertisers through partners, with a CPM (cost per thousand impressions) around $60 and very high minimum commitments—on the order of several hundred thousand dollars. Today, observed CPMs range between $25 and $45, and a CPC (cost per click), mode has been added, with recommended bids around $3 to $5 per click. This range varies by industry, however: Some reports indicate higher bids, in the range of $8 to $18, for more competitive industries like SaaS (Software as a Service) or financial services.

So yes, ChatGPT Ads CPMs are generally higher than those of the Google Display Network (approximately $2 to $5) or Meta (approximately $6 to $10). On the other hand, currently reported CPCs (around $3 to $5) are comparable to those of many industries on Google Search.

And conversion: Is the higher cost worth it?

Comparing costs alone would be reductive. The real question is: "What mindset is the person in when they click?"

Indeed, someone engaging with an AI to compare multiple solutions is often already involved in an active decision-making process. Rather than simply scrolling through a social feed, they are actively seeking an answer. This intent could explain why some early studies report promising conversion rates, even though available data remains limited.

According to internal data reported by Criteo (one of OpenAI's technology partners, based on a sample of hundreds of retailers), relayed by StackAdapt and several industry publications, visitors referred from a conversational AI convert at a rate approximately 1.5 times higher than that of other digital channels. In other words: The click is more expensive, but potentially more profitable—if this ratio holds true for your industry.

As with any emerging channel, caution is warranted: Available data still comes primarily from early technology partners and will need to be confirmed at a larger scale.

The click-through rate (CTR) reported on early campaigns is more modest than that of Google Search—under 1%, compared to an average of 5% to 6% on Google Search. This isn't necessarily a bad sign: In ChatGPT, many users continue their conversation after seeing the ad rather than clicking immediately. This is a different behaviour compared to a traditional search engine, where the click is often the expected action.

What visibility do you have into performance?

As you'll have gathered, advertisers already have access to relatively standard metrics:

• Impressions
• Clicks
• Click-through rate
• Cost per click
• Cost per thousand impressions

Regarding conversion, there are two concrete ways to measure it:

• Directly through OpenAI, with a pixel and a conversions API, launched in May, that track events on your own site
• Via the StackAdapt pixel, if you go through this partner. This is a concrete advantage when you're already a Tink client, since the same tracking pixel you may already be using for your display or native campaigns also captures conversions from ChatGPT, within the same dashboard. You don't need to juggle two separate measurement systems!

ChatGPT Ads: Different from Google Search, but similar to contextual advertising

Between ChatGPT Ads and advertising on Google Search, two distinct logics are at play:

• Search responds to an active intent, since the user is already looking for a brand, a product, or a solution
• ChatGPT Ads, on the other hand, intervenes further upstream, during the consideration, comparison, and exploration phase

These two channels are therefore complementary and serve different objectives.

Moreover, unlike Google Search, advertisers don't know precisely in which conversations their ads are being served.

We trust OpenAI's mechanisms to ensure relevance and protect brands, but this visibility remains limited: Unlike Google Search, you can't see precisely which queries or contexts are involved, which still raises certain transparency and brand safety concerns.

This channel's greatest strengths: Consideration and brand recall

At this stage, ChatGPT Ads seems particularly well-suited to consideration and brand recall objectives. By allowing a brand to appear in a context directly related to the need expressed by the user, this channel can help build consideration and associate the brand with a specific need.

However, you shouldn't expect the same performance in terms of clicks or direct conversions. Cost per click should therefore not be the sole metric used to evaluate this channel, since the person may see the brand, continue their conversation, come back later, and then search for the brand elsewhere. The unique benefit is therefore more about your brand's mental presence in the purchase process, the resulting brand recall, and the association between a need and your brand.

That said, nothing currently allows us to state with certainty that recall will actually be superior to that of other channels. The ad format is still very simple, which potentially limits its impact in terms of brand image. The context is promising, but its actual effect on brand recall remains to be demonstrated. Furthermore, we have no information on ad display frequency: It could therefore be low (resulting in limited recall) or too high (negative impact).

The concrete benefits of advertising on ChatGPT: Why you should pay attention now

• Gain contextual presence at every stage of the journey, not just at the end: Since targeting is based on conversational intent rather than a fixed keyword, a single brand can appear during the initial exploration ("What type of bike would suit me?") as well as during the final comparison ("Which bike should I choose between these two models?")
• Reach a captive audience in decision-making mode : People who use ChatGPT to shop are already engaged in active deliberation—not mindlessly scrolling through a social feed
• Take advantage of a favourable entry window : Spending thresholds have dropped quickly, and competition among advertisers remains low for now. However, this advantage will fade as the channel becomes more mainstream, as was the case for Google Ads and Meta in their early days
• Create a bridge to your organic AI visibility strategy : Testing advertising on ChatGPT today also means learning how your brand is perceived and positioned in generative responses—a competency that is becoming relevant whether you advertise or not.

So, what's our conclusion?

It's still too early to say whether ChatGPT will become an essential advertising channel. However, one thing can already be stated: Conversational interfaces are playing an increasingly prominent role in information and decision-making journeys.

The value of ChatGPT Ads will depend primarily on how this channel is integrated into an overall media strategy. The right approach is to start experimenting with it, clearly define its role in the media journey, set appropriate objectives, and measure results—rather than viewing it as a substitute for other levers.

Like any new channel, ChatGPT Ads still requires a learning period. Best practices around context hints, contextual targeting, and creative writing are still largely being developed. In fact, context hints will likely become a new media competency in their own right, even though we still know little about how to formulate them optimally or which types of descriptions generate the best results.

Moreover, performance measurement is still in its early stages, the cost data currently available is not truly comparable from one advertiser to another, and the link between the delivery context and brand recall remains to be demonstrated. Such grey areas become clearer through testing, not through waiting..

Features, pricing models, and targeting options are evolving rapidly as OpenAI progressively enriches its platform since its commercial launch. This is a situation similar to that of keywords on Google over 20 years ago—we'll need to keep learning with patience and curiosity.

Want to know if advertising on ChatGPT is right for your business?

The Tink team helps organizations evaluate and experiment with new digital levers to make decisions based on data rather than on the trends of the moment.

Contact us to find out how to optimize your overall media strategy.

Mathieu Laliberté

Digital Marketing Specialist
mlaliberte@tink.ca