ChatGPT Sponsored Agents Are Paid Media, Not AEO

OpenAI introduced Sponsored Agents on 16 September 2026: after clicking an ad in ChatGPT, a user can choose to start a clearly labeled conversation with a business-sponsored agent.

For marketers, this creates a useful new paid format. For AEO teams, it creates a measurement trap.

OpenAI says the sponsored conversation is distinct from ChatGPT's independent answer and separate from the original conversation. Its advertising principles also say ads do not influence answers and remain clearly labeled. A paid agent interaction is therefore not an organic mention, an earned citation or evidence that a brand has improved its answer-engine visibility.

The practical response is not to ignore Sponsored Agents. It is to put a hard measurement boundary around them.

What OpenAI actually launched

The launch announcement describes Sponsored Agents as a post-click experience. A user first sees a ChatGPT ad, chooses to engage, and then enters a branded conversation where they can ask follow-up questions and follow a link to the advertiser's site.

OpenAI is testing the format with select advertisers in the United States. The announcement does not describe general self-service availability for Sponsored Agents, a public auction mechanism for the format, or the signals used to determine which agent appears. Those details should be treated as undocumented rather than filled in from assumptions about search ads.

The same release added an Ads Manager plugin for creating, updating and analysing campaigns from ChatGPT Work, AI-assisted creative inside Ads Manager, HubSpot integration and a Shopify app for US merchants. The Shopify app is scheduled to expand internationally on 23 September in markets where ChatGPT Ads are available.

Those are advertising and campaign-management capabilities. None changes OpenAI's stated boundary between ads and independent answers.

Why the boundary matters for AEO

AEO asks whether an answer engine independently names a brand, represents it accurately, or cites one of its pages. Sponsored Agents answer a different question: after paid delivery and an ad click, can a branded conversational experience help the user evaluate an offer?

Both can influence a buying journey. They are not the same exposure.

If a reporting dashboard counts a Sponsored Agent conversation as a ChatGPT mention, three things go wrong:

• Paid reach appears to be earned visibility.
• A campaign launch can look like an AEO improvement even if independent answers never changed.
• The team cannot tell whether budget, content, retrieval or the post-click experience produced the outcome.

OpenAI's own measurement split supports keeping the channels separate. Its Help Center says advertisers receive aggregated reporting such as views and clicks. Ads Manager also supports conversion measurement through the OpenAI Pixel and Conversions API. Those are paid-media measures. They do not report how often an independent ChatGPT answer names or cites the advertiser.

Use a four-layer measurement model

The cleanest implementation keeps four records beside one another without blending their denominators.

1. Paid delivery

Use Ads Manager for impressions, views, clicks, spend and the campaign settings OpenAI exposes. Label every line as paid. If the campaign uses CPM, CPC or conversion optimisation, preserve that objective in the export so unlike campaigns are not compared as though they bought the same outcome.

2. Sponsored conversation

Track the engagement fields OpenAI makes available for Sponsored Agents, but do not invent a proxy when a field is absent. The public launch post does not publish a complete Sponsored Agent reporting schema.

Treat this layer like a conversational landing experience: the user arrived through an ad, the environment is branded, and the questions asked after the click are downstream of paid delivery.

3. Independent-answer visibility

Maintain a fixed prompt set that is run outside the ad and Sponsored Agent flow. Record whether the independent answer mentions the brand, cites a page, describes the offer accurately and names alternatives.

This can be sampled manually. For a repeated cross-engine programme, CiteCue can track ChatGPT and other answer engines against the same prompt set, preserve mention and citation observations, and route supported gaps into a fix queue. Disclosure: AEO Landscape and CiteCue have common ownership; see our full disclosures. CiteCue does not have access to OpenAI's ad auction, Sponsored Agent conversations or internal answer-ranking systems. Its role here is to keep non-paid answer observations separate from campaign reporting.

4. Business outcomes

Use analytics and conversion instrumentation for visits, leads, sign-ups and purchases. Mark ChatGPT Ads traffic as paid at collection time rather than trying to reconstruct it later. Keep independent ChatGPT referrals on a separate channel, and document how unattributed visits are handled.

The four layers can appear on one report. They should never be collapsed into one "AI visibility" percentage.

A clean pre-launch and post-launch workflow

Before launching a ChatGPT Ads campaign, freeze the definitions.

1. Export the current independent-answer prompt sample and record the collection date.
2. Define paid delivery, Sponsored Agent engagement, organic mention, organic citation, referral and conversion as separate events.
3. Add campaign identifiers to the paid links and verify that analytics preserves them.
4. Record the exact landing pages, offer language and campaign start date.
5. Run the same independent prompt sample on a fixed schedule during the campaign.
6. Compare each layer with its own baseline and denominator.

If independent answer visibility changes during the campaign, report the correlation and investigate. Do not claim the spend caused the change. OpenAI explicitly says advertising does not influence its answers, and simultaneous changes to landing pages, product feeds, publicity or demand can move the non-paid sample.

Our open citation-rate protocol provides the sampling discipline for the third layer. The AEO metrics guide explains why repeated observations matter more than a single screenshot.

Audit the agent as paid creative

Sponsored Agents add a quality-control surface that ordinary display ads do not have: the experience can answer follow-up questions. The public announcement does not document how an advertiser supplies, constrains or refreshes the agent's factual material, so teams should obtain those implementation details from their OpenAI account team rather than infer them.

Once access is available, test the paid experience against the same standards used for a landing page and sales conversation:

• Does it state price, availability and limitations consistently with the approved source?
• Does it distinguish documented facts from persuasive language?
• Does it handle unsupported comparisons without inventing competitor weaknesses?
• Does it route regulated, sensitive or account-specific questions appropriately?
• Does the destination page support the claims made in the conversation?

This is brand and campaign QA, not AEO. The distinction makes the work more rigorous, not less important.

What marketers should not claim

Do not describe a Sponsored Agent appearance as an organic ChatGPT citation. Do not combine ad views with prompt-sample impressions. Do not present an increase in paid conversations as a rise in answer-engine share of voice. And do not promise that buying ChatGPT Ads will improve independent answers.

OpenAI has created a paid path from an ad into a branded conversation. That may become a valuable acquisition format. The honest AEO response is to measure it as paid media while continuing to test independent answers on their own terms.

Sources and verification

• OpenAI: Reimagining advertising with AI
• OpenAI: Our approach to advertising and expanding access to ChatGPT
• OpenAI Help Center: Ads in ChatGPT
• OpenAI: New ways to buy ChatGPT ads
• OpenAI: ChatGPT Ads expands across Europe

Sources checked 18 September 2026. This article was researched, drafted and published through AEO Landscape's automated editorial workflow. Product details are attributed to the OpenAI sources above; the four-layer measurement model and audit workflow are AEO Landscape editorial analysis.