A recent discovery reveals that OpenAI's ChatGPT is now serving ads directly within conversational flows, complete with a sophisticated attribution and tracking mechanism. This marks a significant step for OpenAI into advertising revenue, moving beyond subscriptions and API usage. For developers and IT professionals, understanding the underlying mechanics offers valuable insights into the architecture of modern AI monetization and data flow.
What Happened
Research has uncovered the technical details of how OpenAI's ad platform operates, essentially dividing it into two halves: the ChatGPT side and the merchant side. When a user interacts with ChatGPT, the backend injects structured single_advertiser_ad_unit objects directly into the conversation's Server-Sent Events (SSE) stream (chatgpt.com/backend-api/f/conversation). These ad units appear alongside model output, seamlessly integrated into the chat experience.
The ad structure observed is a typed schema, suggesting future potential for multi-advertiser or other complex ad types. Key elements within the ad unit include:
ads_request_id: A unique identifier for the ad request.advertiser_brand: Contains the advertiser's name, URL,adacct_<32-hex>identifier (a stable per-merchant account ID), and afavicon_url.carousel_cards: An array of ad creatives, each with a title, body,image_url, andtarget.target: Specifies the action on click, crucially setting"open_externally": false. This ensures that clicked links open within ChatGPT's in-app webview, allowing OpenAI to observe post-click navigation.
Both brand favicons and ad creative images are hosted by OpenAI on bzrcdn.openai.com, indicating that OpenAI manages the ad creatives rather than directly linking to merchant assets.
Contextual targeting is clearly in play. Observations showed that the same ChatGPT account received different ads across various conversations, each relevant to the chat topic. For instance, a conversation about Beijing trip planning yielded a Grubhub ad for Chinese food delivery, while a discussion about NBA playoffs resulted in an ad for Gametime tickets. This suggests ads are tailored to the immediate conversation, though it's unconfirmed if prior chat history also influences targeting.
{
"type": "single_advertiser_ad_unit",
"ads_request_id": "069e89b3-c038-7764-8000-6e5a193e5f69",
"ads_spam_integrity_payload": "gAAAAABp6Js_<...redacted...>",
"preamble": "",
"advertiser_brand": {
"name": "Grubhub",
"url": "www.grubhub.com",
"favicon_url": "https://bzrcdn.openai.com/cabfae7ead26b03d.png",
"id": "adacct_6984ed0ba55481a29894bb192f7773b4"
},
"carousel_cards": [{
"title": "Get Chinese Food Delivered",
"body": "Satisfy Your Cravings with Grubhub Delivery.",
"image_url": "https://bzrcdn.openai.com/cabfae7ead26b03d.png",
"target": {
"type": "url",
"value": "https://www.grubhub.com/?utm_source=chatgptpilot&utm_medium=paid&utm_campaign=diner_gh_search_chatgpt_kw_traffic_nb_x_nat_x&utm_content=nbchinese&oppref=gAAAA<...>&olref=gAAAA<...>",
"open_externally": false
},
"ad_data_token": "eyJwYXlsb2<...>"
}]
}
The Attribution Chain:
Crucial to the ad platform is a sophisticated four-token Fernet-encrypted attribution chain. Fernet is a symmetric encryption method ensuring both confidentiality and integrity (AES-128-CBC with HMAC-SHA256).
ads_spam_integrity_payload: Sent within the SSE data, this token is used server-side to verify the integrity of ad clicks, preventing fraud.oppref: Present on the click URL, this token is also copied by a client-side tracking SDK (dubbed OAIQ) into a cookie named__opprefwith a 30-day TTL. It acts as a forward attribution token, accompanying subsequent merchant pixel events.olref: Paired withopprefon the click URL, its specific role beyondopprefis not fully detailed in the provided content but likely contributes to the referral tracking.ad_data_token: A base64-wrapped token, also part of the attribution chain, its specific function within the post-click tracking is integrated with the others.
On the merchant side, the OAIQ tracking SDK runs in the visitor's browser, reporting product views and other post-click events back to OpenAI. This, combined with the in-app webview, gives OpenAI comprehensive visibility into the user's journey from ad impression to conversion.
Why It Matters
This development has several significant implications for developers, advertisers, and the broader tech industry:
- Monetization Strategy: For OpenAI, integrating ads directly into ChatGPT opens up a powerful new revenue stream, diversifying its business model beyond API access and premium subscriptions. This could allow for more robust investment in R&D and scaling AI infrastructure.
- Ad Tech Evolution: This represents a novel application of advertising within conversational AI interfaces. Developers interested in ad tech can study this model for injecting dynamic, contextually relevant ads into real-time streams (like SSE) and for building sophisticated, multi-stage attribution systems.
- Data and Privacy: The comprehensive tracking, including in-app webviews and persistent cookies (
__oppref), highlights the significant data collection happening. Developers need to be aware of such mechanisms, especially when integrating AI services into their applications, and consider the privacy implications for end-users. Users will also increasingly need to understand how their interactions contribute to ad targeting. - Developer Tooling & Integration: As OpenAI further develops its ad platform, it might offer tools or APIs for advertisers to manage campaigns, creative assets, and performance. This could create new opportunities for developers to build complementary services or integrate their products with ChatGPT's advertising ecosystem.
- Competitive Landscape: OpenAI's entry into AI-powered advertising could intensify competition with existing ad giants (Google, Meta) and other AI companies exploring similar monetization strategies. It also sets a precedent for how conversational AI platforms might be monetized in the future.
- Content Hosting: OpenAI hosting advertiser creatives (
bzrcdn.openai.com) demonstrates a deep integration into the ad delivery pipeline, offering potential benefits in terms of performance and control but also placing more responsibility on OpenAI for content moderation and compliance.
What To Watch
Going forward, developers and IT leaders should keep an eye on several fronts:
- Expansion of Ad Formats: Will OpenAI introduce more complex ad units, perhaps interactive or rich media formats, beyond the observed
single_advertiser_ad_unit? - Targeting Sophistication: How will ad targeting evolve? Will it integrate deeper user profiles, purchase history, or cross-platform data, potentially raising further privacy concerns?
- Developer APIs for Ad Management: Will OpenAI expose APIs that allow developers and agencies to programmatically manage ad campaigns within ChatGPT, similar to existing ad platforms?
- User Response and Regulation: The introduction of ads into a conversational AI will inevitably lead to user feedback. How will OpenAI balance monetization with user experience, and how might regulatory bodies (especially concerning data privacy) react to this new form of advertising?
- Integration with Other OpenAI Products: Could this ad platform extend to other OpenAI offerings, such as DALL-E or custom GPTs, creating a broader advertising network within the OpenAI ecosystem?
The move to monetize ChatGPT through advertising is a significant development, underscoring the growing maturity and commercial ambitions of conversational AI. Understanding the technical underpinnings is crucial for navigating this evolving landscape.