OpenAI is testing Sponsored Agents. After seeing an ad in ChatGPT, a user can choose to open a clearly labeled conversation with a business-sponsored agent, ask product questions, compare fit, and visit the business website when ready. That conversation is separate from ChatGPT’s independent answers and the user’s original chat. At announcement, the test was limited to select advertisers in the United States. Read OpenAI’s details on the ad experience and business tools.

OpenAI official mark
OpenAI official mark; click to open the announcement.

How this differs from a display ad

A conventional ad mainly sends a person to a website. Sponsored Agents turn an ad click into a conversation that can handle follow-up questions. OpenAI’s example is a shopper asking whether an advertised table fits a room, how many people it seats, and how to care for its finish. A business agent may explain product details, but users should be able to tell they are speaking with a merchant representative rather than receiving independent ChatGPT advice.

AI tools for advertisers

Advertisers can use the Ads Manager plugin in ChatGPT with natural-language prompts to create, update, and analyze campaigns. A website or brief can become a draft campaign, and the plugin can help interpret performance. Ads Manager can also suggest copy and images based on a landing page and campaign objective. Advertisers can review and edit suggestions before adding them. AI copy customization and translation are opt-in; check prices, terms, and disclaimers in each language before launch.

OpenAI also announced HubSpot and Shopify integrations so businesses can connect an ad account, view performance, and follow up on leads in their existing CRM or ecommerce tools. Fewer interface changes still require careful review of catalog data, customer information, attribution, and permissions.

A merchant pilot

  1. Start with one product category and a small budget. Keep prices, stock, specifications, returns, and FAQs current.
  2. Define what the brand agent must not answer and when it should hand a person the conversation. Test ambiguous requests, stale inventory, and poor product fit.
  3. Report ad conversations separately from organic answers. Measure conversation starts, qualified questions, click-to-purchase conversion, refunds, and misleading-claim complaints.
  4. Have product, legal, and marketing teams review ad labeling, AI-edited copy, and consent for customer data before launch.