Stenberg's reporting frames the leaked StackAdapt deck as revealing that ChatGPT prompts — where users disclose health symptoms, proprietary code, and personal anxieties — are being categorized and sold to programmatic ad buyers. He highlights the conversational format actively encourages deeper disclosure than search ever did, making this a qualitatively different kind of data monetization.
The editorial argues that people treat ChatGPT like a confessional, sharing things they'd be embarrassed to even Google. The nature of conversational AI actively encourages this disclosure, and now that intimate data is being packaged into targeting categories for DSP buyers — a fundamentally different privacy calculus than traditional search ads.
StackAdapt's deck explicitly positions ChatGPT inventory as 'the highest-intent digital surface ever created,' arguing that detailed natural-language prompts provide richer intent signals than any search query. They are selling this access at CPMs in the $15-50+ range, reflecting the premium value of real-time conversational intent data.
The editorial notes that OpenAI recruited Shivakumar Venkataraman, Google's former VP of Ads engineering, and placed monetization under CFO Sarah Friar from Nextdoor. It argues these hires tell the story better than any press release — you don't recruit Google's ads VP to merely explore advertising, you recruit them to build a full ads business.
The editorial directly compares OpenAI's claim that raw prompts aren't shared with advertisers — only aggregated topic categories — to the same distinction Google drew 20 years ago. It argues this is 'technically true in the same way,' implying the distinction erodes over time as the ad infrastructure matures and targeting becomes more granular.
Adweek's Mark Stenberg published a leaked sales deck from StackAdapt, a Toronto-based demand-side platform valued at $2.1 billion, that lays out exactly how advertisers can buy ad placements inside ChatGPT conversations. The targeting mechanism is called "prompt relevance" — ads are matched to the topic and intent of what a user is asking the AI in real time.
The deck positions ChatGPT inventory as "the highest-intent digital surface ever created," and it's hard to argue with the logic. When someone types a detailed natural-language query about, say, migrating their Kubernetes cluster, they've just handed advertisers a richer intent signal than any search query ever could. StackAdapt is selling this at CPMs reportedly in the $15-50+ range — search advertising prices for what used to feel like a private conversation.
OpenAI's ad push is being led by Shivakumar Venkataraman, formerly Google's VP of Ads engineering, and monetization strategy falls under CFO Sarah Friar, who joined from Nextdoor in 2024. The hiring tells the story better than any press release could: you don't recruit Google's ads VP to explore advertising. You recruit them to build an ads business.
The core issue isn't that ChatGPT has ads. It's the nature of the data being monetized.
People treat ChatGPT conversations like confessionals. They ask about health symptoms they're embarrassed to Google. They paste proprietary code and ask for debugging help. They describe relationship problems, financial anxieties, legal situations. The conversational format actively encourages disclosure — that's the whole point of a chat interface. Now that disclosure is being categorized and sold to programmatic ad buyers through standard DSP infrastructure.
OpenAI's defense is that raw prompts aren't shared with advertisers — only aggregated topic categories. This is the same distinction Google drew 20 years ago, and it's technically true in the same way that "we don't sell your data, we sell access to you" is technically true. The practical reality: OpenAI analyzes your private prompts to build targeting segments, and advertisers buy access to those segments. Whether the advertiser sees your exact words is beside the point when the commercial outcome is identical.
The programmatic infrastructure angle is arguably more concerning than OpenAI's own ad plans. StackAdapt is a DSP — a platform that lets *any* advertiser bid on inventory across *any* publisher in their network. Once ChatGPT ad inventory is available through standard programmatic pipes, it becomes commodity inventory. Every ad-tech middleman, every retargeter, every data broker in the programmatic supply chain gets a piece. This isn't a curated, "thoughtfully integrated" ad experience. It's real-time bidding on conversational intent data, running through the same infrastructure that serves the banner ads you ignore on recipe blogs.
The Hacker News discussion (259 points) was predictably sharp. The dominant thread: "This was always the endgame." OpenAI's trajectory from nonprofit research lab to $157 billion valuation company made ad monetization a mathematical certainty. Sam Altman's earlier dismissals of advertising plans now read as either naive or deliberately misleading. Multiple commenters drew the Google parallel — another company that started with "don't be evil" aspirations and ended up building the most sophisticated advertising surveillance system in history.
Here's where it gets concrete for practitioners. Once a language model's deployment is funded by advertising, the model's operators have a financial incentive to maximize ad-relevant engagement. This doesn't mean ChatGPT will start recommending products in its responses tomorrow. It means the system now has a structural reason to prefer conversations that are longer, more frequent, and in higher-CPM topic categories.
Consider the implications for developer workflows:
Code assistance quality. If developer-tool advertisers pay premium CPMs for prompt-relevant placements, there's an incentive to keep developers asking follow-up questions rather than providing the most efficient answer. The metric that matters shifts from "solved the user's problem" to "generated high-value impressions."
Free tier vs. paid tier divergence. ChatGPT Plus ($20/month) and Pro ($200/month) subscribers presumably get ad-free experiences. But 200M+ weekly active users on the free tier become the ad inventory. Over time, this creates pressure to gate the best model capabilities behind the paywall while keeping the free tier optimized for ad engagement. If you're a developer relying on the free tier for quick lookups, expect the experience to degrade.
Data handling. Every prompt you send now passes through an ad-targeting classification layer. If you're pasting code snippets, architecture diagrams, or error logs into ChatGPT, that content is being analyzed not just to generate a response but to determine which ad category your conversation falls into. Your company's security team should be aware of this additional data processing purpose.
The practical response depends on how you use ChatGPT today.
If you're an individual developer using the free tier for ad-hoc coding questions, this is your cue to evaluate alternatives. Anthropic's Claude and locally-run models (Llama, Mistral) don't have advertising business models. The open-weight ecosystem has reached the point where a quantized 70B model running on a decent GPU provides genuinely useful code assistance without sending your prompts anywhere.
If your organization has a ChatGPT Enterprise or Team subscription, you're likely insulated from ads — but read the updated terms of service carefully. The question isn't whether enterprise prompts are used for ad targeting today. It's whether the data processing infrastructure being built for the free tier creates new risk vectors for enterprise data.
If you're building products on the OpenAI API, the immediate impact is minimal — API calls are separate from the consumer ChatGPT product. But watch the incentive structure. As advertising becomes a larger share of OpenAI's revenue, product decisions across the board will be influenced by what maximizes ad inventory value.
The ad-supported AI assistant is now the industry default, not the exception. Google's Gemini has ads through Google's existing infrastructure. Meta's AI chatbot had advertising from day one. OpenAI was the last major holdout among the big consumer AI products, and that's over. The companies betting against ads — Anthropic, Apple — are now the contrarians, and their business models look increasingly like a competitive advantage for users who value privacy. For developers, the lesson from 20 years of ad-supported search is clear: the product that respects your attention and your data eventually costs money. The one that's free eventually costs something else.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.