How OpenAI ad tests could reshape post-click web
▼ Summary
– OpenAI is testing a ChatGPT ad format where clicking an ad opens a business-specific AI agent conversation instead of a traditional landing page.
– The system uses three components: automated business profiling from website crawling, configurable business agents with custom instructions and data sources like product feeds and MCP tools, and campaigns that direct users into agent conversations.
– The technology is built on the same foundation as Custom GPTs, and the ads are currently available to a limited group of advertisers in ChatGPT Ads Manager.
– This model shifts digital advertising from a click-opening-a-webpage paradigm to the conversation itself being the destination, enabling immediate answers, product recommendations, and lead qualification.
– The end-user experience has not been widely observed yet, so it’s unclear how these ads will appear or how prominently they’ll be surfaced in ChatGPT.
OpenAI is quietly testing a new advertising format inside ChatGPT that could turn the classic click-through model on its head. Instead of sending users to a landing page, these ads would open a dedicated, business-specific AI conversation right inside the chat interface.
The shift is subtle but potentially seismic. For decades, the ad-to-webpage pipeline has been the backbone of digital marketing. A user clicks, a browser loads, and the brand makes its pitch. OpenAI’s experiment replaces that final step with something more conversational: an AI agent trained on the company’s own data, ready to answer questions, recommend products, and even capture leads without the user ever leaving the chat window.
The system appears to work in three distinct phases. First, ChatGPT crawls the advertiser’s website to build an automatic business profile. This profile captures common customer questions, support details, and general company context. Next, advertisers can customize a business agent using their own instructions. That agent merges the scraped profile with additional inputs like product feeds, Model Context Protocol (MCP) tools for real-time business data, and custom lead-generation forms. Finally, campaigns are launched not to a URL, but directly into a conversation with that agent.
The underlying architecture looks similar to what powers Custom GPTs, suggesting this is less a one-off test and more a strategic direction.
Why this matters. The traditional ad funnel is linear: impression, click, webpage. This model collapses that journey. The conversation itself becomes the destination. A user could get immediate answers, book an appointment, troubleshoot a product issue, or get pre-qualified for a sale before ever seeing a traditional website.
The feature appears to be live now in the ChatGPT Ads Manager, though only for a limited set of advertisers. The end-user experience hasn’t been widely observed yet, so it’s still unclear how these ads will render inside ChatGPT or how much placement prominence they’ll receive.
Still, the infrastructure points to a clear intention. OpenAI is building toward chatbot-native advertising, where a click summons an AI sales representative rather than a static page.
The discovery was shared by Entrepreneur Juozas Kaziukėnas, who spotted the new ad type and posted about it on LinkedIn.
(Source: MarTech)
