What are Sponsored Agents in ChatGPT ads?

The short version

Sponsored Agents are a new ChatGPT ads format that lets a user open a chat with an AI representative of a business, launched from an ad. OpenAI's entire published description of the format is four sentences long: what it is, that it is alpha only, that it is not open to general advertisers, and that OpenAI is not accepting early access requests. Everything else in this guide is either drawn from OpenAI's wider ads documentation, applied to what a Sponsored Agent probably inherits, or clearly marked as our own view.

What OpenAI has actually published

Here is the complete substance of OpenAI's Sponsored Agents help article, in full:

Sponsored Agents are an advertising format that lets people chat with an AI representative of a business in ChatGPT. People can open a conversation from an ad to ask questions about the business's products or services.

Three further facts sit alongside that definition. Sponsored Agents are currently available only to selected advertisers in a limited alpha test. A ChatGPT user may see a Sponsored Agent even if the format is not available for their own campaigns, meaning exposure to users is ahead of advertiser access. And OpenAI is not accepting early access requests at this time, stating only that it will announce more details about availability and how to participate when it is ready to expand access.

That is the entire public record as of this page's verified date. Any claim beyond those four points, about pricing, setup, measurement, eligibility criteria or a launch date, is not something OpenAI has stated, and we are not going to invent one to fill the gap.

What we do not know

• How it is billed. Standard ChatGPT ads bill on impressions, clicks, or a conversion event. OpenAI has not said which model, if any, applies to a Sponsored Agent conversation, or whether a conversation itself is a billable or measured event.
• Who builds the agent. Whether an advertiser supplies its own content, a script, a set of approved answers, or something else has not been described.
• What guardrails apply. OpenAI's general principle is that ads do not influence ChatGPT's own answers, and that advertisers cannot shape, rank or alter ChatGPT's responses. Whether a Sponsored Agent counts as part of "ChatGPT's response" or sits outside that boundary as advertiser controlled content has not been clarified.
• Eligibility criteria for the alpha. OpenAI names no category, spend level or account history requirement for being selected.
• A timeline. No date, no quarter, no "coming soon" marker has been published.

What is likely to carry over from the standard ad format

OpenAI's wider ads documentation sets rules that apply to advertising in ChatGPT generally, not to one format, so it is reasonable, though not confirmed for Sponsored Agents specifically, to expect these to apply:

Treat this table as an informed expectation, not a confirmed spec. We will update it the moment OpenAI publishes format specific rules.

How a Sponsored Agent probably differs from the standard unit

The standard ad is a static block: a title, copy, image and a link out to a landing page. A Sponsored Agent, by OpenAI's own description, is a live conversation that opens inside ChatGPT itself rather than sending the user away to a website. That is a meaningfully different job for the advertiser. A landing page is written once and refined with analytics. A conversational representative has to hold up across an open ended set of questions, in whatever order a user asks them, without OpenAI's chat model itself being the thing answering, since OpenAI is explicit that ads do not influence ChatGPT's own responses. Whatever technical boundary keeps a Sponsored Agent's answers separate from ChatGPT's own model has not been described publicly.

What good preparation looks like, in our view

Nobody can build a Sponsored Agent today because nobody outside the alpha group has access. What a business can do now is build the assets a conversational format would need regardless of the eventual mechanics:

• A clean, current FAQ. Sponsored Agents will only be as good as the source material behind them. A business whose product details, pricing and policies are scattered or outdated is not ready for a conversational ad format even once it opens.
• Context hints written properly now. OpenAI's context hint framework, what you offer, who it helps, when it is useful, already forces the kind of structured thinking a conversational agent will likely need. See writing context hints for ChatGPT ads.
• A working standard ads account. Alpha programmes for new ad formats tend to draw from advertisers already active and in good standing on the base platform, though OpenAI has not confirmed this for Sponsored Agents specifically. An account with no campaign history is a weaker position to request access from once OpenAI opens the door.
• A process for keeping product information current. If your team cannot answer "how would we keep this accurate" for a static ad, a conversational one raises the stakes on getting that answer right.

Why this sits behind an alpha, not a beta

OpenAI's terminology is worth noticing. Ads Manager itself, the platform every standard advertiser uses, is labelled beta: broadly available to eligible self service advertisers, with capabilities still evolving. Sponsored Agents sit a step further back, described as a limited alpha available only to selected advertisers. That is a narrower, earlier stage than the beta the rest of this site's guides describe, and it lines up with how OpenAI rolled out the base platform: a closed pilot from February through April 2026, before Ads Manager Beta opened self service in May. If Sponsored Agents follow a similar path, alpha would be expected to precede a broader beta before anything resembling general availability, though OpenAI has not confirmed that sequence for this format, and has given no indication of timing for any of those stages.

Why we are not speculating further

Some competitor content on this topic fills the gap with confident predictions: what it will cost, when it will launch, what verticals it will suit best. We are not doing that here. OpenAI has published four sentences about this format. A guide that pads that out with invented specifics would read as authoritative and be wrong. If you came here looking for a cost estimate or a launch date, the honest answer is that OpenAI has not published one, and nobody outside its alpha group is in a position to state one as fact.

Questions worth asking once access opens

1. How is a conversation billed, and what counts as the billable or measured event: an opened chat, a message sent, a completed exchange, or something else?
2. What content or script does the advertiser supply, and who reviews it before it goes live, given that every other ChatGPT ad format goes through review before serving?
3. What happens when a user asks the agent something outside its brief? A standard ad cannot be asked an open ended question. A conversational one will be.
4. What measurement does the advertiser receive? Standard ads report aggregated views and clicks. A conversation implies richer signal, but OpenAI's stated principle is that advertisers only see what a user sends them directly, so it is unclear whether aggregate conversation analytics will be offered on top of that.
5. Does the format carry its own eligibility bar, separate from standard ads account verification, particularly for regulated categories such as healthcare or financial services where OpenAI already applies stricter criteria to the standard format?

We will answer each of these the moment OpenAI does, and not before.

Where this could matter most, if it works as described

Framed strictly as our own reasoning from the stated definition, not a platform claim: a format that lets a user ask follow up questions before leaving ChatGPT would matter most for businesses whose sales process is naturally a conversation rather than a single decision, the kind of purchase where a static ad and landing page force a user to either commit or leave. Company formation, property, private education and considered ecommerce purchases are the categories where a UAE business we work with typically fields the most pre-purchase questions before converting elsewhere, which is why we think this format is worth tracking even while it is unavailable. That is a judgement call about which businesses would benefit most, not a claim about how OpenAI intends to allocate alpha access, and OpenAI has given no indication of which categories it is testing with.

UAE and Gulf specifics

Sponsored Agents are not tied to a market in OpenAI's published description, so a UAE advertiser is in the same position as any other: no route to request access, no published timeline, and exposure as a user possible before exposure as an advertiser. If access does open with a country gate the way standard Ads Manager access did, a UAE business would need its advertising entity registered in an available country, the same rule that already governs standard self service access. See ChatGPT ads in the UAE for that entity requirement, and the Arabic and English language handling that a conversational format would presumably also need to support, given that OpenAI's ad language selection already considers a user's location and language settings for the standard format.

What to do while you wait

Run the standard ad format now. It is self service, available in the UAE, and everything you learn about your offer, your audience and your context hints will transfer directly to a Sponsored Agent if and when your account is invited into that alpha. Start with what ChatGPT ads are and how they work, and track any change to Sponsored Agents availability on our ChatGPT ads news and rollout timeline, which we update whenever OpenAI publishes something new.

Related