Milena Traikovich has built a career on turning data into demand, helping businesses navigate the complex world of analytics and performance optimization to nurture high-quality leads. As OpenAI enters the advertising arena with a price tag rivaling an NFL broadcast, we sat down with our Demand Gen expert to dissect the real-world implications of this seismic shift. In our conversation, Milena unwraps the daunting economics behind a $60 CPM, explores the strategic pivot to a new “intent economy,” and grapples with the critical challenges of data opacity and attribution. She also offers a stark look at the governance required to operate in this new space and a glimpse into a future where brands must learn to market not just to humans, but to their AI agents.
OpenAI is reportedly charging a premium $60 CPM for ads, far above typical social media rates. What specific ROI metrics must a brand achieve to justify this cost, and how does that calculation differ from a sub-$20 CPM on another platform? Please share a hypothetical example.
To even consider a $60 CPM, you have to throw out your standard top-of-funnel metrics and focus exclusively on high-value, bottom-of-funnel conversions. This isn’t about cheap clicks or broad awareness; it’s about surgically precise lead generation. The entire calculation shifts from Cost Per Mille (CPM) to a ruthless focus on Customer Acquisition Cost (CAC) and Lifetime Value (LTV). Let’s imagine a B2B SaaS company selling a compliance tool. On a platform with a sub-$20 CPM, they might get thousands of impressions, many irrelevant, to secure one qualified demo. With ChatGPT, the promise is that the user has already articulated a complex need, like “find me a compliance software that integrates with my existing stack for the financial industry.” That lead is so pre-qualified that the sales cycle could be drastically shorter. So, if your LTV for a new client is $100,000, paying a higher initial cost to acquire them directly through this ultra-targeted channel could result in a much healthier CAC-to-LTV ratio than a broader, less efficient campaign ever could. You’re not just buying an ad; you’re buying a pre-vetted introduction at the exact moment of need.
The shift to an “intent economy” suggests ads can target users after their needs are fully articulated. How does this change the creative approach for advertisers, and what specific steps should they take to adapt their messaging for this post-intent environment?
It’s a fundamental change in mindset from interruption to affirmation. In the old “attention economy,” your ad creative needed to scream to be heard over the noise. It had to grab attention, create a need, and pull a user away from what they were doing. In this new “intent economy” that experts like Calvin Scharffs describe, the user has already done the heavy lifting. They’ve told the AI exactly what they want. So, the first step for marketers is to stop shouting and start listening. Your creative should feel less like an ad and more like the perfect answer, a helpful next step. Instead of a flashy banner, think of a concise, value-driven message that confirms you have the solution. The messaging needs to be direct, transparent, and aligned with the context of the conversation. For example, if a user asks for “the best hiking boots for rocky terrain under $200,” your ad shouldn’t just show a boot; it should say, “Our XT-5 model: Rated #1 for rocky trails, built for durability, and priced at $189.” It’s about being the solution, not just another option.
Marketers face uncertainty about the specific targeting parameters and performance data available in conversational AI platforms. What are the biggest risks of committing a large budget with this data opacity, and what initial, smaller-scale tests could a brand run to mitigate those risks?
Committing a budget of nearly $1 million upfront, as is being reported, without clear data is an enormous gamble. The biggest risk is flying blind. You’re essentially trusting an entirely new, black-box algorithm with a massive chunk of your budget. Without understanding the targeting parameters, as analyst Nicole Green points out, you have no way to optimize, no way to learn what works, and no way to truly justify the spend to your CFO. You could be burning through cash on poorly targeted placements and have no data to explain why. To mitigate this, brands should push for pilot programs with much smaller, controlled budgets. An initial test could involve a highly specific, niche product or service where intent is unambiguous. For example, a brand could run a campaign targeting only conversations around a specific software integration problem it solves. The goal wouldn’t be massive scale, but rather to establish a baseline. Can we measure a lift in direct site traffic or branded search queries originating from users in the test cohort? It’s about proving the model’s efficacy on a small scale before even thinking about a seven-figure commitment.
In a conversational AI, an ad might appear alongside an organic recommendation for the same product. How can marketers effectively distinguish between the influence of the paid placement versus the AI’s recommendation, and what new attribution models might be needed to measure true campaign impact?
This is the attribution nightmare that keeps performance marketers up at night. The line between an organic recommendation and a paid ad becomes incredibly blurry, and traditional last-click models are rendered completely useless here. To tackle this, we need to move towards more sophisticated attribution models, likely a combination of multi-touch attribution and incrementality testing. Marketers will need to establish control groups—users who are eligible to see the recommendation but are intentionally not served the paid ad. By comparing the conversion rates between the group that saw both the ad and the recommendation versus the group that only saw the recommendation, we can start to measure the true “lift” or incremental impact of the paid placement. Furthermore, clear and consistent labeling is non-negotiable. If users feel tricked, trust in both the brand and the platform will plummet. The future of attribution in this space will rely on a foundation of transparency and controlled experimentation to prove the ad’s value.
For regulated industries, advertising within an AI conversation presents unique compliance challenges. What specific governance framework should a marketing team establish with its legal and compliance departments before launching a campaign, and what are the key red flags to watch for?
For any company in a regulated industry like finance or healthcare, this is a minefield. The first step in building a governance framework is to bring legal and compliance into the room before a single dollar is spent—this is a point Nicole Green from Gartner has emphasized, and it’s critical. The framework must include a rigorous review process for all ad creative and messaging to ensure it cannot be misinterpreted as advice, especially when appearing next to an AI’s output. A key red flag is any ambiguity in how user data is handled. The marketing team needs to get explicit, written confirmation from the platform about what conversational data is used for targeting and what is shared. Another red flag is the lack of control over ad adjacency. What if your ad for a financial product appears in a conversation where the AI is giving flawed financial guidance? The brand could face reputational damage or even legal liability by association. The framework must outline a clear process for monitoring these adjacencies and an escalation plan for when things go wrong.
As AI agents become “machine customers” that transact for humans, marketing strategies must evolve. What is the first practical step a brand should take today to make its product information “AI-readable,” and how will this prepare them for a future of both human and machine journeys?
The most crucial first step is to treat all of your public-facing product information as a structured dataset. Right now, a lot of marketing content is designed for human emotion and persuasion. To prepare for machine customers, you need to supplement that with data that is clean, structured, and unambiguous. This means going beyond beautiful marketing copy and implementing detailed product schemas, clear specifications, and transparent pricing and inventory data that an API can easily parse. Think like a machine: does your website clearly state your product’s dimensions, compatibility, ingredients, or API endpoints in a standardized format? As Caroline Giegerich noted, your brand’s reputation—reviews, third-party validation—also becomes part of this dataset. By structuring this information now, you’re not just optimizing for today’s search engines; you’re building the foundation that future AI purchasing agents will use to evaluate and buy your products. This makes you discoverable and trustworthy to both the human researching you today and the machine buying for them tomorrow.
What is your forecast for the evolution of conversational AI advertising over the next three to five years?
Over the next three to five years, I predict we’ll see a dramatic split in the market. Initially, the high price point will mean only major brands with huge budgets can afford to experiment, and we’ll see a lot of trial and error. But as the technology matures and compute costs hopefully decrease, we’ll see the emergence of more accessible, self-serve platforms for conversational ads, much like we saw with search and social. The real evolution, however, will be in integration. Advertising won’t just be a static placement; it will become a dynamic, interactive part of the conversation. Imagine an ad that functions as a mini-app within the chat, allowing a user to configure a product, ask follow-up questions, and even complete a purchase without ever leaving the conversation. We’re moving from placing ads in conversations to making the ads part of the conversation. The brands that will win will be those who master this seamless, value-added interaction, effectively turning their advertisement into a service.
