People don’t open ChatGPT the way they open Instagram. There’s no doom-scrolling, no passive browsing. They open it because they need something: a plan, an answer, a way out of a problem.
That changes everything about how advertising works inside it.
OpenAI recently began testing ads in ChatGPT in the US, partnering with major agencies including WPP, Publicis Media and Dentsu. Ads are currently shown to logged-in users aged 18 and above on free or Go subscriptions. And unlike any platform before it, targeting is driven primarily by what users are doing in the conversation right now, not who they are.
If you’re a founder or marketing manager trying to figure out whether this matters to your business, this post breaks down exactly how ChatGPT advertising works, how it differs from what you’re used to, and what you should be thinking about before it becomes mainstream.
ChatGPT advertising refers to sponsored content shown within conversations on the ChatGPT platform, but it’s nothing like a banner ad or a Google search result.
When a user asks ChatGPT “what laptop should I buy for my small business?”, an ad for a laptop brand could appear as part of the interface. The trigger isn’t a keyword. It’s the intent behind the conversation.
According to early information from OpenAI’s ad test, targeting signals include:
This makes ChatGPT advertising fundamentally different from any platform that came before it.
For decades, digital advertising ran on a simple premise: figure out who someone is, and guess what they want.
If you’re a 24-year-old living in Amsterdam, you might see ads for gym memberships. If you’re a 45-year-old with a mortgage, you might see ads for home insurance. Platforms like Google and Meta became extraordinarily good at building detailed profiles; age, gender, location, browsing history, purchase behaviour, and selling access to those profiles to advertisers.
It worked. Until it started not working.
The problem with demographic targeting is threefold:
In a static format like a search results page, these limitations are manageable. In a conversational platform like ChatGPT, they become a real problem.
Behaviour-based targeting skips the profile entirely and focuses on the signal you already have: what the person is asking about right now.
👉 Instead of: “This user is a 30-year-old marketing manager in Utrecht.” The system focuses on: “This user is currently asking how to set up a marketing budget for a product launch.”
That’s a completely different kind of signal, and a much more useful one.
In traditional search advertising, intent is inferred from keywords. Someone searching “email marketing tool” is probably in the market for one. That’s been the foundation of Google Ads performance for 25 years.
ChatGPT takes this further. Conversations are multi-step, exploratory, and contextual. A user might ask:
That entire conversation is a rich signal. A relevant ad for a lead generation tool, shown at the right moment, could feel genuinely useful rather than disruptive.
That’s Coby’s core belief about advertising on any platform: the right message, to the right person, at the right moment. Behaviour-based targeting in AI is the closest the industry has ever come to actually delivering that.
This isn’t all upside. There are three real challenges worth understanding.
Sensitive conversation contexts. ChatGPT users regularly ask about mental health struggles, financial crises, relationship problems, and career anxiety. Placing an ad for a debt consolidation service inside a conversation about financial stress isn’t neutral, it can feel exploitative. OpenAI will need clear policies about which conversation contexts trigger ads, and which don’t.
The trust problem. ChatGPT is a tool people use for help. The moment users start to feel like responses are shaped by advertiser interests rather than their own, even if that’s not true, trust in the platform collapses. Transparency in labelling is non-negotiable here. Users need to clearly see what’s an ad and what’s an answer.
Limited reporting. Unlike Google Ads or Meta Ads, OpenAI currently offers no detailed conversion reporting. Advertisers can’t track what happened after an ad appeared in a conversation. That’s a significant gap, and it means ChatGPT advertising is still far behind more mature platforms when it comes to performance measurement.
For context: at Coby, we treat tracking as the foundation of every campaign we run. Without reliable conversion data, you can’t optimise anything. That’s as true for ChatGPT ads as it is for Google or Meta, and right now, the data infrastructure simply isn’t there yet.
The current version of ChatGPT advertising has a high entry barrier. Early reports suggest minimum spend requirements that effectively lock out smaller advertisers for now.
But the principle behind it, reaching users based on what they’re actively trying to solve, is extremely well-suited to smaller, more focused businesses.
A tutoring service that appears when someone asks “how do I help my child prepare for their exams?” A budgeting tool that surfaces when someone asks “how do I manage cash flow as a freelancer?” That level of relevance is difficult to achieve on traditional platforms without a significant budget and audience size.
As the platform matures and opens up, SMEs with tightly defined products and clear use cases will likely outperform large brands that rely on mass awareness. The specificity of intent-based targeting rewards niche, not scale.
Every time a new ad platform emerges, there’s a rush to be first. Sometimes that pays off. Often, it doesn’t, because the fundamentals of good advertising don’t change just because the channel does.
At Coby, we ask the same questions regardless of platform: Is the tracking reliable enough to measure what matters? Does the targeting align with where your customer actually is in their journey? Are we reaching people based on real signals, or guesswork?
ChatGPT advertising isn’t ready for most of our clients right now. The reporting gaps are too significant for us to recommend it with confidence. But the underlying model, serving ads based on intent rather than identity, is exactly the direction the industry needs to go.
When the infrastructure catches up, this will be a genuinely interesting channel. We’ll be watching closely.
ChatGPT advertising is an early experiment, not a mature channel. But, the principle it’s built on, behaviour and intent over demographics and guesswork, is the most important shift in digital advertising targeting in years.
For SME founders and marketing managers, the practical takeaway right now is simple: don’t rush in. Watch how the platform develops, what reporting capabilities emerge, and whether the ethical guardrails around sensitive content hold up.
When it’s ready, the businesses that understand their customer’s intent, not just their demographics, will have the advantage.
Want to make sure your current advertising is built on the right foundations first? Book a free audit →
ChatGPT advertising is sponsored content shown within conversations on the ChatGPT platform. Unlike traditional ads, targeting is based on what users are currently discussing rather than demographic profiles.
Google Ads targets based on search keywords and user data. ChatGPT advertising targets based on conversational intent, what the user is actively asking about in real time, making it more contextual but currently less measurable.
Not yet, practically speaking. The current test involves large agency partners and high minimum spend thresholds. Broader access for SMEs is likely to follow as OpenAI scales the programme.
It can be. Behaviour-based targeting focuses on the current conversation rather than a historical data profile, which reduces the feeling of being “followed around” the web. However, how conversation data is stored and used over time is a separate question that OpenAI hasn’t fully addressed.
For most SMEs, not yet. The lack of conversion reporting makes it difficult to measure ROI. Focus on building solid tracking and performance foundations on mature platforms first, then evaluate ChatGPT advertising as reporting improves.
Relevance to the exact conversation topic, clear labelling as an ad, and a message that supports the user’s goal rather than interrupting it. Ads that feel like helpful suggestions perform better than those that feel like intrusions.