The next important ChatGPT ad may not send anyone to a website. OpenAI is introducing a more conversational advertising experience through Sponsored Agents. In its Ad Tools Terms, OpenAI defines ChatGPT Sponsored Agents as advertiser-sponsored conversational experiences where users can interact with an AI-generated representative of a business, product or service.
Instead of treating the ad click as the end of the media experience, Sponsored Agents could turn that click into the beginning of a conversation. A user could potentially ask questions, compare options, understand a product or service and move toward the next action without immediately leaving ChatGPT.
That makes this more than another ad format. OpenAI is effectively testing whether the post-click experience itself can become conversational.
The timing matters. ChatGPT Ads has already moved beyond its early US test, with self-service access expanding into markets including India as part of a broader international rollout. Our earlier breakdown of the ChatGPT Ads expansion into India and more than 40 countries covered that infrastructure shift. Sponsored Agents point to the next question: what happens after an advertiser wins the click?
The ad click could become the landing page
Current OpenAI documentation describes the standard ChatGPT ad as a placement below a response containing an advertiser name, headline, description, image and landing page. Advertisers can measure impressions, clicks, spend, CTR, CPC, CPM and conversions.
ChatGPT Sponsored Agents change that flow in the current pilot. Digiday reports that advertisers can attach a branded business agent to an ad, so clicking the CTA takes the user into a branded chat within ChatGPT instead of immediately sending them to the advertiser’s website.
OpenAI’s own Ad Tools Terms, published on August 24, define a Sponsored Agent as an advertiser-sponsored conversational experience where users interact with an AI-generated representative of a business, product or service.
That produces a very different funnel.
| Area | Standard ChatGPT ad | Sponsored Agent pilot |
| Click destination | Advertiser landing page | Branded chat in ChatGPT |
| Main interaction | Read, browse, convert | Ask, compare, clarify, then act |
| Advertiser asset | Copy, image, destination | Ad plus branded AI agent |
| Measurement | Existing click and conversion reporting | Full Sponsored Agent reporting has not been publicly detailed |
| Current status | Available through ChatGPT Ads in supported markets | Select-advertiser pilot |
OpenAI has already been adding more conventional performance controls, including conversion-focused buying and stronger measurement. Our analysis of ChatGPT Ads conversion bidding showed how quickly the platform has been adding the basics advertisers expect.
The Sponsored Agent idea moves in the opposite direction. Instead of making ChatGPT Ads look more like Google or Meta, it starts making the format look more native to ChatGPT.
ChatGPT Sponsored Agents change what “creative” means
A normal paid media team can spend hours debating a headline, an image and a CTA. With ChatGPT Sponsored Agents, that may become the easy part.
The advertiser may also need to think about what happens after someone asks questions such as “Which plan works for a 50-person team?”, “Does this integrate with Salesforce?”, “Can you compare these two products?” or “Can I talk to sales?” That turns ad experience design into conversation design.
OpenAI’s Ad Tools Terms make the responsibility fairly explicit. Advertisers remain responsible for a Sponsored Agent’s content, configuration, actions and outputs, as well as the materials supplied to build it. Those materials can include product information, pricing, inventory, scripts, brand guidelines and business logic.
For marketing teams, that creates four practical jobs:
- Product information has to stay accurate. Old pricing or stale feature claims become more dangerous when an agent can repeat them in real time.
- Sales and marketing scripts need alignment. The agent should not promise something the sales team later has to walk back.
- Human handoffs need rules. Some questions should end in a page visit. Others should route to sales or customer service.
- Agent quality becomes part of media quality. A strong ad followed by a weak conversation is still a weak campaign.
This is where the ad format starts overlapping with the operational questions we covered in OpenAI Presence and enterprise AI agents. Running an agent in front of customers requires more than a prompt. It requires approved data, actions, escalation rules and ongoing review.
Conversational advertising is becoming a platform pattern
OpenAI is not the first company to put conversation closer to an ad. Google’s Business Agent for Leads inside Search ads gives users a conversational path from an ad, while Meta Business Agent across messaging channels pushes businesses toward AI-assisted conversations in WhatsApp, Instagram and Messenger.
The interfaces are different, but the commercial idea is similar. Instead of Ad → website → form → follow-up, the funnel can become Ad → conversation → qualification → action.
That is why AI-native advertising and broader conversational advertising deserve attention beyond the novelty of talking to an ad. For ecommerce, the agent could answer product questions before checkout. For services, it could clarify fit before a lead form. For B2B SaaS, a branded AI agent could potentially handle early questions around use cases, integrations, plans or implementation before handing a prospect to sales.
Potentially is doing a lot of work there. OpenAI has not published evidence showing that the format improves conversion rate, qualified lead rate, revenue or CAC. There is no public Sponsored Agent benchmark yet.
The website disappearing creates a measurement problem
ChatGPT Sponsored Agents could reduce friction, but they also remove something marketers understand reasonably well: the website session. A website gives teams pageviews, navigation paths, form events, remarketing signals and analytics data. If more evaluation happens inside ChatGPT, marketers need a different way to define progress.
A click alone will be far too weak. For B2B teams, useful signals might eventually include a qualified conversation, requested demo, product recommendation, sales handoff or completed purchase. OpenAI has not publicly detailed which Sponsored Agent events advertisers will receive or how they will connect to Ads Manager reporting.
That makes the principles in our B2B marketing attribution guide even more relevant. The optimisation event has to connect to pipeline, not simply produce a prettier platform metric.
Privacy needs the same caution. For standard advertising, OpenAI says advertisers do not receive users’ private ChatGPT conversations, memories, names, emails or other personal details. But marketers should not assume that existing ad reporting documentation fully answers how data generated inside a Sponsored Agent conversation will be handled. A detailed public product specification has not yet been released.
ChatGPT Sponsored Agents still have more unknowns than benchmarks
The pilot is interesting precisely because so much remains unresolved. OpenAI has not publicly disclosed Sponsored Agent pricing, auction mechanics, broad eligibility, rollout timing or detailed measurement. Digiday also reported that OpenAI did not respond to its request for comment on the format.
Policy could matter just as much as product design. OpenAI continues to define where ads can appear and which advertiser categories qualify, particularly around sensitive contexts. The company’s ad policies were updated again on September 10.
Teams in finance, healthcare and other regulated industries should therefore pair any future agent experiment with the same compliance discipline covered in our analysis of OpenAI’s changing ad policies for health and finance.
The format deserves testing when access broadens. It does not deserve blind budget yet.
What marketing teams should watch next
Three developments will tell us whether ChatGPT Sponsored Agents become a serious advertising format rather than an interesting demo. First, watch measurement: advertisers need to know which conversational actions can be tracked, optimised and connected to downstream revenue. Second, watch control: brands will need clear ways to manage approved claims, product data, agent actions, escalation and human review. Third, watch economics: keeping someone inside ChatGPT may increase engagement, but that only matters if those conversations produce better business outcomes than sending the same person to a strong landing page.
A longer conversation is not automatically a better conversion.
OneMetrik Takeaway
The most interesting part of ChatGPT Sponsored Agents is not that an ad can talk back. It is that the boundary between media and the conversion experience is starting to disappear.
Paid media teams have spent years optimising what happens before the click. Landing page and CRO teams handled what happened after it. A Sponsored Agent puts both jobs into one conversational surface.
For OneMetrik, the test would be simple: do not judge this format on clicks or chat starts. Judge it on qualified conversations, sales handoffs, pipeline and revenue.
If OpenAI can connect conversational intent to measurable business outcomes without weakening privacy or advertiser control, click-to-chat ads could become one of the first formats that genuinely belongs inside an AI assistant. Until then, the idea is ahead of the evidence.