ChatGPT Sponsored Agents review (2026): an honest verdict on OpenAI's conversational ad format — high intent and clean labeling, but an early US-only test.
Sponsored Agents is the most interesting thing OpenAI has shipped in advertising: a labeled, walled-off conversation with a brand's AI agent captures more of the sales dialogue inside ChatGPT than any card can. But at announcement it's an early US-only test with select advertisers, there are no published results, no disclosed buying model, and the agent is only as good as the content you feed it. Worth preparing for now; not something most brands can buy or judge yet.
This review scores ChatGPT Sponsored Agents as an advertising format — the thing a marketer would eventually buy — as of OpenAI's September 16, 2026 announcement. It is not a review of ChatGPT the assistant, and it can't be a review of proven performance, because the format is an early test with a small group of advertisers in the United States. What it can do honestly is rate the design: how promising the format is, where it's strong, and where it's still raw.
The scores rate what will decide whether this format is worth a media budget: how much buyer intent it captures, how deep the conversation can go, how it's labeled and walled off, how available it is, what it demands of your content, and how legible the results are likely to be. Where the design is genuinely strong — a high-intent, conversational surface with clean labeling — it scores well. Where it's unproven — no benchmarks, no disclosed pricing, US-only, and heavily dependent on the content behind it — the scores say so.
The most useful takeaway isn't the composite. Sponsored Agents point clearly at where high-intent discovery is heading, and the work that makes one good — deep, current, on-brand product content — is the same work that earns you an organic ChatGPT mention today. So the smart move is to prepare regardless of when you can buy in, not to wait for access.
ChatGPT Sponsored Agents is a conversational ad format OpenAI began testing on September 16, 2026. When an ad is relevant, a user can tap it and open a clearly labeled conversation with a business-sponsored AI agent — asking follow-up questions and clicking through to the business's website when ready. OpenAI keeps that conversation distinct from ChatGPT's own independent answers and separate from the user's original ChatGPT conversation, so the agent is clearly the brand talking, not the neutral assistant. Its illustrative example is a shopper asking a furniture brand's agent about a table's dimensions, seating, and care. The format shipped alongside broader ad tooling: natural-language campaign management in the Ads Manager, AI-suggested copy and imagery generated from your landing pages, opt-in text customization that adapts headlines and descriptions to a conversation and translates ad copy, and OpenAI's first CRM and ecommerce partners, HubSpot and Shopify. At announcement, Sponsored Agents were limited to select advertisers in the US, and OpenAI did not publish a buying model, pricing, or performance data specific to the format. It is distinct from the existing labeled sponsored ad card, which is a one-way unit rather than a conversation.
Right now, Sponsored Agents fits almost no one as a buyable channel — it's a test with select US advertisers, so for most brands the honest fit is "prepare, don't buy." As a format to plan for, it suits businesses whose purchases hinge on real questions an agent could answer well: considered products with specs, comparisons, compatibility, and care details, where a back-and-forth genuinely helps a shopper decide. It's a strong conceptual fit for brands that already invest in deep, current product content, because the agent is a front-end over exactly that. It's a weak fit for anyone expecting a turnkey channel with predictable cost today, or for brands with thin product content, since the agent can only be as helpful as the material behind it.
| Dimension | Score | Why |
|---|---|---|
| Buyer intent captured | 4.3 / 5 | A conversation opened from a relevant ad is deep mid-funnel intent — the shopper is actively evaluating and asking to be sold. |
| Conversational depth | 4.0 / 5 | Answering follow-ups and objections inside ChatGPT captures far more of the sales dialogue than a static card, when the content supports it. |
| Labeling & brand safety | 4.1 / 5 | Clearly labeled and walled off from ChatGPT's own answers and the user's original chat, so the brand voice can't be mistaken for the assistant's. |
| Availability & access | 1.8 / 5 | An early test with select advertisers in the US only; most businesses cannot run one at announcement. |
| Content dependency / setup burden | 2.6 / 5 | The agent answers only from the content you provide, so a good conversation is a real content-production project, not a quick campaign. |
| Reporting & proven results | 2.2 / 5 | No performance data or benchmarks published for the format; early adopters are the baseline. |
| Ad-creation tooling | 3.6 / 5 | Natural-language campaign management, AI-suggested creative from your pages, and opt-in text customization and translation are a solid operator kit. |
| Ecosystem integrations | 3.8 / 5 | HubSpot as first CRM partner and Shopify as first ecommerce partner (free app) ease campaign, lead, and catalog operations. |
| Cost transparency | 2.3 / 5 | No buying model or pricing disclosed for Sponsored Agents specifically at announcement; unknowable to forecast. |
The honest answer on pricing is that OpenAI did not publish one for Sponsored Agents at announcement — no buying model, no CPC or per-conversation rate, no minimum. The existing ChatGPT Ads card is bought self-serve on a cost-per-click or reach basis, but the agent format is a separate, earlier-stage test, and it would be a guess to assume it inherits the same economics. So "is it fairly priced" is unanswerable today; treat any number you see as unconfirmed until OpenAI states it.
The cost that is knowable is the one that doesn't show up on an invoice: the content behind the agent. A Sponsored Agent that answers a shopper's real questions requires deep, current, structured product content — specs, comparisons, compatibility, care, returns — and producing and maintaining that is the actual investment, whether or not you ever pay for the placement. Budget for the content first; it's the part that determines whether the format works and the part that pays off on the organic side regardless.
The pragmatic read: you can't forecast Sponsored Agents as a channel yet, and you shouldn't try. What you can do is make the content that a future agent — and ChatGPT's unpaid answers — will draw on, so that when access and pricing arrive, the expensive part is already done.
| Use case | Fit | Why |
|---|---|---|
| Considered purchases with real buyer questions | Strong | A conversation that settles dimensions, compatibility, and comparisons is exactly where the format shines. |
| Brands with deep, current product content | Strong | The agent is a front-end over your content, so an existing content investment converts directly into a good conversation. |
| Most SMBs wanting to buy in today | Weak | It's an early US-only test with select advertisers; access isn't generally available. |
| Ecommerce running on Shopify and HubSpot | OK | The first-partner integrations ease operations, but the format itself is still a limited test. |
| Advertisers needing predictable cost from day one | Weak | No disclosed pricing or performance data means no way to forecast the channel. |
| Brands with thin or stale product content | Weak | A poorly-fed agent deflects to the website, wasting the intent it captured. |
| Teams preparing content for AI discovery now | Strong | The content that makes a future agent good is the same content that earns organic ChatGPT mentions today. |
To be clear, Kompozy is not an ad format and won't compete with Sponsored Agents for a media budget. If your goal is to run a conversational ad inside ChatGPT, that's OpenAI's product, and Kompozy doesn't sell placements. They sit at different points in the same funnel.
Where Kompozy matters is the thing this format depends on and doesn't provide: the content. A Sponsored Agent is a front-end over your product knowledge, and it can only answer as well as the material you've published — while the organic mention next to it is earned by that same material. Kompozy produces that content at volume from one Persona Brief: direct-answer blog articles, brand-exact spec carousels and infographics, quote graphics, and avatar-narrated shorts that answer buyer questions on camera, all in one voice and published across the eight social platforms plus blog and email. So the framing is complementary — OpenAI supplies the conversational slot; Kompozy supplies the content that makes both the paid agent and the free mention worth having.
As a buyable channel, it's not a decision most brands can make yet — at announcement it was an early US test with select advertisers and no published pricing or results. As a direction, it's worth preparing for now, because the deep, current product content that makes an agent good is the same content that earns organic ChatGPT mentions today.
OpenAI did not publish a buying model or pricing for the Sponsored Agents format at its September 16, 2026 announcement. The separate ChatGPT Ads card is bought on CPC or reach, but you shouldn't assume the agent inherits that. Confirm current terms on OpenAI's advertising pages before planning a budget.
The existing ChatGPT Ads unit is a one-way labeled card at the bottom of an answer. A Sponsored Agent is a two-way, clearly labeled conversation with a business's AI agent, kept separate from ChatGPT's own answers. The card competes on earning a click; the agent competes on answering a shopper into a decision.
The content behind it. The agent is a front-end over your product knowledge, so it can only answer from what you've published — specs, comparisons, compatibility, care, returns. Deep, current, structured content produces precise conversations; thin content produces vague ones that deflect to the website.
No. At announcement it was limited to select advertisers in the United States, so most businesses cannot buy one. The practical step for everyone else is to build the product content an agent — and ChatGPT's organic answers — would draw on, and check OpenAI's pages for expanded access.
AI ad-creation tools — natural-language campaign management, AI-suggested creative from your landing pages, and opt-in text customization and translation — plus HubSpot as its first CRM partner and Shopify as its first ecommerce partner, with a free ChatGPT Ads app for merchants.
Inventory the questions that decide a purchase in your category and make sure each has a clear, current, on-brand answer across the surfaces a model reads. That readiness work pays off on the organic side immediately and makes the paid agent effective the moment you can run one. A content engine like Kompozy produces and maintains that footprint at scale.
See ChatGPT Sponsored Agents (OpenAI) vs Kompozy comparison → · Get Started →