The landscape of digital advertising shifted significantly in May 2026 when OpenAI finally opened its self-serve ad platform to the public. For years, marketers viewed the conversational interface of ChatGPT as a walled garden, but the arrival of a native ad system has created a “wild west” environment where early adopters are currently carving out their territory. To navigate this new frontier, we are joined by Milena Traikovich, a seasoned Demand Gen expert who specializes in high-quality lead nurturing and performance optimization. With her deep background in analytics, Milena has been at the forefront of testing these new placements, moving past the initial $200,000 minimum commitments of the pilot phase to help brands of all sizes leverage the $25-a-day entry point. Our discussion explores the technical nuances of “context hints” over traditional keywords, the surprising engagement data comparing AI chats to search engines, and the strategic pivot required when targeting the massive Free and Go tier user base.
The conversation today centers on the practical realities of launching a ChatGPT campaign, from the granular details of the relevance-weighted, second-price auction to the specific constraints of the 18-and-older demographic requirement. We delve into why a $3 CPC bid is often the hard-coded “floor” for delivery and how the lack of demographic or keyword targeting forces advertisers to rethink their creative messaging. Milena shares insights from real-world case studies, including the performance gap between keyword-heavy prompts and conversational questions, providing a roadmap for those ready to move beyond Google and Meta.
Since ads currently reach only the 18-and-older demographic on free and low-tier accounts, how does this shift the strategy for brands that usually target premium, high-income users who likely pay for ad-free tiers?
Targeting the Free and Go tiers requires a fundamental shift in how we perceive the “value” of a user, as those paying for ChatGPT Plus or higher tiers are effectively invisible to this ad platform. For brands used to a high-income persona, this means we must focus on the “moment of interest” rather than the user’s subscription status, capturing them during their research phase when they are using the tool for free. We have to realize that even high-earning professionals often use the free version for quick, logged-out sessions or casual inquiries, so our creative must be sharp enough to stop a scroll at the bottom of a conversation. It is a bit of a balancing act because if your product only appeals to the top 1% of spenders, you might find the current reach frustratingly limited. However, the 18-plus demographic is massive, and by focusing on the “context hints” of the conversation, we can find those users when they are specifically asking about high-value topics like home renovation or business software.
The platform replaces traditional keyword targeting with a feature called “context hints.” How should an advertiser approach this freeform box to ensure the AI matches their ad to the right conversation?
The context hint box is perhaps the most exciting and nerve-wracking part of the setup because it is a blank slate that doesn’t follow the rigid “exact match” or “phrase match” rules we grew up with in Google Ads. In our testing, we found that dumping a list of keywords—like “lumber yard near me” or “buy cedar planks”—actually performed better and achieved a lower average CPC than trying to be overly “conversational” with long-winded questions. You have the option to write a carefully crafted prompt, such as describing a DIY homeowner in San Antonio looking for decking materials, but the AI seems to respond very effectively to clear, thematic pillars. It feels like a hybrid between a search query and a creative brief, where you are guiding the system’s matching engine rather than setting hard boundaries. The key is to avoid being too narrow; since there is no “negative keyword” list yet, your context hints need to be descriptive enough to attract the right intent without accidentally pulling in irrelevant traffic.
With average CPCs currently sitting between $2 and $5, and a hard-coded delivery threshold of $3, how can smaller businesses justify this spend compared to more established channels?
The $3 minimum bid is a significant barrier for many SMBs, especially when they see a warning that their ad might not even deliver if they bid lower. This isn’t a dynamic assessment based on competition yet; it feels more like a fixed entry fee for the platform’s early days, which can be a bitter pill to swallow for those used to $0.50 clicks on social media. However, you have to weigh this against the $25 daily budget minimum, which is incredibly accessible compared to the six-figure commitments required just a few months ago during the pilot. We justify the cost by looking at the intent: a user asking ChatGPT how to build a deck is in a much more active state of “doing” than someone mindlessly scrolling through an Instagram feed. Even if the initial CPC feels high, the potential to be the only “Sponsored” result at the end of a long, helpful AI explanation provides a level of authority that is hard to replicate elsewhere.
We see a stark difference in engagement time—17 seconds for ChatGPT Ads versus 41 seconds for Google Ads. What does this tell us about the mindset of a user interacting with an AI versus a search engine?
The drop to 17 seconds of engagement time is a sobering reality check that reminds us that ChatGPT users are often on a mission to get a specific answer and move on. In a search engine, users are accustomed to clicking through multiple tabs and comparing results, whereas the AI conversation provides a sense of “completion” that might make them less likely to linger on a landing page. This suggests that ChatGPT Ads are currently functioning more as a high-intent “pulse” than a deep-dive research tool, requiring our landing pages to be lightning-fast and incredibly relevant to the chat topic. We noticed in our case study that the conversion rate was also significantly lower—0.21% compared to 3.71% on Google—which highlights that we are catching users in a conversational flow rather than a final buying window. To succeed here, your ad needs to feel like the natural “next step” in the conversation, offering a solution that the AI might have just described in theory.
The reporting for ChatGPT Ads is described as feeling like the “early days” of Performance Max, with very little transparency into what actually triggers an ad. How do you optimize a campaign when you can’t see the conversation data?
Optimizing in a “black box” environment forces you to lean heavily on the few levers you actually control: the creative assets, the destination URL, and the geographic targeting. Since we cannot see the actual keywords or conversation themes that triggered the ad, we have to create separate campaigns for different regions if we want any granular geographic data, as reporting only segments by country by default. You have to treat the platform with a certain level of trust, similar to how we handled the early rollout of automated bidding, while being meticulous with your own third-party tracking. We strongly recommend using OpenAI’s pixel to track button clicks and order revenue, but pairing that with UTM parameters in GA4 is non-negotiable if you want to see the truth behind the 1,500 users hitting your site. Without that outside data, you’re essentially flying blind, relying on the platform’s “relevance-weighted” auction to do the heavy lifting for you.
OpenAI’s ad policies can be self-contradictory, particularly regarding legal services. What is your advice for businesses in “gray area” industries trying to navigate these eligibility hurdles?
The policy landscape is currently a bit of a moving target, with the official documentation saying one thing while the actual changelog and live ads suggest another. For instance, while the policy page might say legal services are not permitted, we have seen ads for lawyers running successfully since May 2026, which creates a confusing environment for compliance-heavy industries. My advice is to always check the most recent changelog before launching and be prepared for a manual review process that might feel inconsistent. If you are in a restricted category like financial services or alcohol, you need to be extremely careful with your favicon and headline to ensure you aren’t flagged by an automated system that is still learning the nuances of the policy. It is better to start with very “safe” creative and then incrementally test the boundaries of what the system allows as you establish a positive account history.
For businesses with a large inventory, the new product ad format includes star ratings and prices. How “clunky” is the setup process for these feeds currently?
The product feed process is definitely in its “beta” stages of frustration, and we’ve encountered multiple glitches where data doesn’t sync correctly or images are rejected for no apparent reason. You have to submit a square image of at least 256 by 256 pixels, and even then, the way the AI truncates a 50-character headline can make your product look unprofessional if you aren’t careful. The update on July 24 that added prices and star ratings was a huge step forward, but it adds another layer of complexity because if your feed isn’t perfectly formatted, those crucial trust signals won’t appear. It is a time-consuming endeavor compared to the relatively seamless feed uploads on Google or Meta, but for e-commerce brands, the effort is worth it to appear as a “suggested product” at the moment someone asks for a recommendation. You just have to go in with the expectation that you will need to troubleshoot the technical integration multiple times before it runs smoothly.
What is your forecast for the future of AI-driven advertising?
I believe we are rapidly approaching a point where AI-driven advertising will move from being a “sidebar” experience to becoming the primary way we discover new brands, eventually rivaling the scale of traditional search. Within the next two to three years, I expect OpenAI to solve the “transparency gap” by providing advertisers with anonymized “intent clusters” that show us the general themes of conversations our ads are winning. We will likely see the $3 CPC floor drop as the auction becomes more efficient and more tiers of users are opened up to advertising, perhaps even including limited ads for premium subscribers who opt for a lower monthly fee. The real game-changer will be when these ads become interactive themselves, allowing a user to ask questions about the sponsored product directly within the chat interface without ever leaving the conversation. It is a complete reimagining of the marketing funnel, and those who master the “context hint” logic today will be the ones leading the industry when AI search eventually overtakes the traditional blue-link results we’ve relied on for decades.
