OpenAI Is Testing Ads That Talk Back — And Rebuilding the Marketing Job While It Does

OpenAI announced on Wednesday that it is testing Sponsored Agents in ChatGPT: after clicking an ad, a user can start a labelled conversation with an agent paid for by the business behind it. The company says the test is running with select advertisers in the United States, and that the sponsored conversation is held separate from ChatGPT's own answers and from the thread the user started. In the same post, OpenAI put ChatGPT Ads inside HubSpot, its first CRM partner, and inside Shopify, its first ecommerce partner, with the Shopify app going international on September 23. The consensus reading is that search advertising has arrived in chat. That understates it.

The unit being sold is the conversation

Banner advertising sells an impression. Search advertising sells an intent. A Sponsored Agent sells the dialogue that follows the click — the follow-up questions, the objections, the reasoning a buyer does out loud before deciding. OpenAI's own example is a shopper asking whether a dining table fits their space and how to care for its finish. That exchange used to happen in a showroom with a salesperson, or in a browser tab the buyer controlled. It now happens inside a system funded by one side of the transaction.

OpenAI is explicit that the sponsored conversation is distinct from ChatGPT's independent answers. Take the company at its word and the problem does not go away. The buyer still leaves the neutral surface and enters a paid one, carrying the same trust posture they had ten seconds earlier. Labelling tells you where the boundary is. It does not tell you what you lose by crossing it.

The second change is quieter. OpenAI is launching opt-in AI-powered text customization that adapts an advertiser's headlines and descriptions to the context of a conversation and translates copy into the user's preferred language. The creative is no longer fixed at the moment of the buy. It is assembled against what you just said.

The marketing job is being rebuilt from the middle

Read the rest of the announcement as a labour document. Advertisers can now use natural-language prompts in ChatGPT to create, update and analyse campaigns through the Ads Manager plugin, turning a website or a brief into a live campaign. Ads Manager itself now drafts suggested copy and imagery from the landing page and the campaign objective. HubSpot users can create ads, track performance and follow up on leads without leaving their CRM.

Every one of those is a task that currently sits on someone's job description. Campaign setup. Copy variants. Performance summaries. Lead follow-up drafting. None of them is the whole role, which is exactly how this always starts — the tasks are absorbed first, the title survives one more planning cycle, and the headcount question arrives afterwards. The part of the job that survives is the part that decides what the campaign is for, what claim the brand is allowed to make, and which channel deserves the budget at all.

Your Next Move

Audit where your category's buying conversation now happens. Open ChatGPT and ask it, as a customer would, to compare your product against two named competitors. Write down what it says and what it omits. If a competitor is buying the agent slot next to that question, you are being argued against in a room you have never entered.

Split your marketing tasks into brief-writing and brief-execution. List what your team shipped last month in those two columns. Execution — variant copy, campaign builds, performance decks — is what OpenAI just automated inside the tools your team already uses. Move your own hours toward the first column before the budget conversation does it for you.

Set a rule for sponsored agents before you need one. Decide now that product research inside a paid conversational surface gets verified against a source you chose. The label is doing real work, and the moment to decide how much weight you give it is before a vendor is answering your questions for you.