OpenAI ads changed shape on September 16, 2026. Until yesterday, buying placement in ChatGPT meant buying a labelled slot next to an answer. Now it can mean buying a conversation.
OpenAI opened a pilot for Sponsored Agents: click an ad, and a branded agent opens a separate, labelled chat you can talk to. Angi and Wayfair are named. Pricing, targeting and rollout are not.
Most coverage of this is going to be about the ad product. The part worth your attention is quieter. Every tool your team uses to track AI visibility watches organic answers. A Sponsored Agent is not an organic answer.
What OpenAI ads are now
OpenAI ads are labelled paid placements shown to logged-in Free and Go users in supported countries, running on systems separate from the model that writes answers. As of September 16, 2026, they also include Sponsored Agents: a branded conversational agent a user can open from an ad click, held in its own thread apart from ChatGPT's independent reply.
Measurement coverage · September 2026
Five ChatGPT surfaces. Your tooling covers two of them.
What you can observe yourself, and what an AI visibility platform reports.
ChatGPT cites your page in its independent response
Every GEO platform is built for this row
A click lands on your site with a chatgpt.com referrer
GA4 and server logs catch it; most tools infer it
ChatGPT-User pulls your page mid-conversation
Only visible in your own edge logs
Impressions and clicks inside ChatGPT Ads Manager
No prompt-level data on what triggered the ad
A competitor holds a branded conversation with your buyer
No instrument in the category observes this
Assessment of Profound, Peec AI, Otterly, Scrunch and PromptWatch as documented by each vendor in September 2026. Sponsored Agents were announced on September 16, 2026 and no vendor has published coverage of the format.
What actually shipped on September 16
The format is real, the pilot is small, and the disclosures are thin. Here is what is on the record.
OpenAI's own announcement pages have returned 403 to us for fourteen consecutive days, so the company's wording below is quoted as reported by The Next Web, not as something we fetched from OpenAI. The confirmation that carries legal weight comes from somewhere better: an issuer press release.
Angi (NASDAQ: ANGI) published a release on September 16 confirming it is among the first brands in the pilot. A listed company saying this in a press release is a far stronger source than a trade blog, because it carries securities exposure.
That last row is the honest state of it. There is no published CPM, no CPC model, no advertiser count, no country list beyond the United States, and no general availability date.
Two named brands and no published price tells you where the surface is heading, and nothing about what to spend.
The detail that matters is in Angi's release, not OpenAI's
Angi already had a ChatGPT app. It launched earlier in 2026 and it worked the way apps work: the user chooses to add it.
The sponsored agent is different, and Angi says so plainly. In the company's words, "the existing Angi app is something homeowners actively choose to add in ChatGPT, while the Angi-sponsored agent can appear within relevant conversations as a sponsored experience that users may choose to engage with."
Read that twice. The product change is not that brands can build agents. Brands could already do that.
The change is that a paid agent can surface inside a conversation the user never started with that brand.
Opt-in became opt-out, and the price of entry is an ad buy.
Five things a Sponsored Agent changes that a banner slot did not
A labelled ad slot competes for attention. An agent competes for the conversation. Those are different problems.
Change #1: The ad now holds the objection handling, and you never see the objections
A static placement gets a click or it does not. An agent gets asked about price, contract length, security review, and the competitor the buyer is also considering.
Whatever gets asked in that thread belongs to the advertiser and OpenAI. It does not reach the rest of the market, and it does not reach you.
Change #2: Your competitor can appear in a conversation about your category
The Angi wording is the whole story. Relevance triggers the placement, not brand intent.
If your category is the topic, a competitor who bought the placement can open a branded thread with a buyer who was researching the category generically.
Change #3: The format sits beside an answer that already cites sources
This is where paid and organic collide. Our own 63-day study of 90,132 AI answers found ChatGPT cited a source in 92.5% of them. The full numbers are on our AI search statistics page.
An answer that cites sources and a sponsored thread that does not are now rendered a few hundred pixels apart. The buyer is reading both.
Change #4: Paid presence does not buy organic presence
StubGroup, an agency that runs these campaigns with client budgets, puts it plainly in its ChatGPT advertising guide, updated September 16: ads do not influence the organic results ChatGPT gives users, and running ads will not increase organic mentions of your business.
Paid placement buys the conversation. It does not buy the citation.
That matches what we see in client audits. A brand can be spending on ChatGPT placement while being absent from the answer a buyer gets when they simply ask the category question.
Change #5: Consumer tolerance for this is not high, and it is measurable
Semrush surveyed 2,338 US adults in July 2026 and found that 41.63% of consumers dislike chatbot ads. Of those, 66.05% said ads make them doubt the integrity of AI responses. The same study found 65.01% of AI users have at least partially swapped Google for chatbots in product research.
So the audience is moving to the surface and is suspicious of paid content on it at the same time. A conversational ad format is a bet that a good agent conversation overcomes that suspicion. It might. Nobody has data either way yet.
Do you know what a buyer sees when they ask ChatGPT about your category?
We run the category prompts, record which brands and sources ChatGPT returns, and show you where your organic presence ends and a competitor's begins.
Book an AI Visibility AuditThe blind spot: what your AI visibility tool cannot see
Here is the structural problem, and it is worth being precise about it.
What an AI visibility platform measures:
- •Which brands appear in organic AI answers for your tracked prompts
- •Which domains and URLs get cited in those answers
- •How those rates move week to week
- •Sentiment and position inside the organic response
What a Sponsored Agent conversation is:
- •Opened from an ad click, not from an answer
- •Held in a separate thread, explicitly distinct from the independent answer
- •Labelled as sponsored, so it is not a citation
- •Invisible to prompt-based sampling, because no prompt produces it
Run those two lists side by side and the conclusion is uncomfortable. Profound, Peec AI, Otterly, Scrunch and PromptWatch all sample organic answers. None of them documents any coverage of a paid conversational surface, because the surface is one day old and nobody has built for it.
Your dashboard will not move on the day a competitor starts holding branded conversations with your buyers.
We flagged this term before the product existed. OpenAI's advertising policy document, updated July 31, 2026, defined sponsored agents as conversations users hold with an AI representative of an advertiser's business, and the company declined to elaborate. That was one unexplained line in a policy PDF. Thirty-five days later it is a pilot.
Being early to the definition does not give us data on volume or impact. It gives us the better question to ask: what will I never see if a competitor buys one?
Your own instruments have the same gap
We log every AI bot request to cite.solutions at the edge. That gives us a clean read on the answer-time layer: in a 29-day window we found ChatGPT-User was 41% of every citation-capable request hitting the site.
That instrument is good at what it does and useless here. A sponsored agent conversation happens inside ChatGPT. If it never fetches your pages, it never touches your logs.
The same is true of the advertiser's own view. StubGroup notes that ChatGPT gives advertisers no information about the conversations or signals that triggered their ads. Even the brand paying for the placement does not get prompt-level data.
Five steps to take before Sponsored Agents leave the pilot
You cannot buy the format outside the US pilot, and there is no price to evaluate. So the work is preparation, not procurement.
Step 1: Decide whether paid conversational placement is even on your roadmap
Write down the honest answer for your category. High-consideration B2B software with a six-month cycle and a security review is a different proposition from home services with a same-week hire.
If the answer is no, you still need steps 2 through 5, because a competitor's yes changes your environment.
Step 2: Baseline your organic presence before the paid layer thickens
Run the category prompts now and record which brands and sources ChatGPT returns. Do it before the surface gets noisier, so you have a clean reference point.
This is the ordinary work of a brand mention audit across AI platforms, and it gets more valuable the moment paid formats start competing for the same screen.
Step 3: Ask your AI visibility vendor what they plan to do about paid surfaces
The honest answer today is "nothing yet." Ask anyway, and ask it in writing.
A vendor who tells you their organic share-of-voice number is unaffected by paid conversational placement is telling you something true and incomplete. Our guide to choosing an AI visibility tool covers what to press on.
Step 4: Separate your paid ChatGPT reporting from your GEO reporting
If you already run ChatGPT ads, keep the two ledgers apart. Ad clicks are not citations, and blending them produces a number that flatters the campaign and hides the organic gap.
Our guide to running ChatGPT ads for B2B SaaS covers the media side; the question of whether B2B SaaS should run conversion ads at all is a separate decision.
Step 5: Make the organic answer strong enough to survive the comparison
This is the only lever that works in every scenario. A buyer reading a sponsored thread on one side and a cited answer on the other is running a trust comparison whether they know it or not.
If your pricing, security and implementation pages are the ones ChatGPT cites in that answer, the paid thread next to it is competing with your own evidence. If they are not, it is competing with nothing. A managed GEO agency can run that work continuously if your team does not have the bandwidth for a monthly cycle.
What not to conclude from this
A few guardrails, because the temptation to over-read a launch is strong.
- •Two named advertisers is a pilot, not a channel. It may stay small.
- •No pricing has been published, so any CPM or CPC figure you see attributed to Sponsored Agents is invented.
- •OpenAI's existing ad product runs on Free and Go tiers. Do not assume Sponsored Agents inherit that restriction, because nobody has said so.
- •We have no data on reach, volume or effect. The gap we describe is structural, not measured.
The claim here is narrow and I think it holds: a new commercial surface opened next to the answer, and the instruments the category sells cannot observe it.
Your AI visibility report covers organic answers. What about everything beside them?
Cite Solutions audits how your brand appears across ChatGPT, Gemini, Claude, Perplexity and AI Overviews, and tells you plainly which surfaces our measurement can and cannot reach.
Book a Discovery CallFAQ
Does ChatGPT have ads?
Yes. OpenAI shows labelled ads to logged-in Free and ChatGPT Go users in supported countries, and paid tiers stay ad free. Ads run on systems separate from the model that generates answers, and advertisers receive aggregate performance data rather than access to chats.
What are Sponsored Agents in ChatGPT?
Sponsored Agents are a paid format announced on September 16, 2026. A user who clicks an ad can open a clearly labelled conversation with a business-sponsored agent, held separately from ChatGPT's independent answer. It is a test with select US advertisers, with Wayfair and Angi named.
Can you buy a Sponsored Agent right now?
Not unless you are in the pilot. OpenAI has described it as a test with select advertisers in the United States and has published no pricing, targeting options, advertiser criteria or general availability date.
How much do OpenAI ads cost?
For the existing ad format, practitioner guides report a minimum daily budget of $25 per campaign, with geographic targeting down to postal code and custom audiences requiring a minimum matched-user threshold. No pricing has been published for Sponsored Agents.
Do ChatGPT ads improve my organic AI visibility?
No. Advertisers and OpenAI both state that ads do not influence ChatGPT's independent answers, and running ads does not increase how often the model mentions your business. Organic citation is earned through the content the model retrieves, which is a separate workstream.
Bottom line
Sponsored Agents are a small pilot with two named brands and no published price. Treated as a media opportunity, there is nothing to act on yet.
Treated as a measurement question, there is. A paid conversational surface now sits next to the organic answer, and every instrument the AI visibility category sells points at the organic side only.
Baseline your organic presence this month, keep your paid and organic ledgers separate, and ask your vendor what happens to their numbers when the thing they cannot see starts mattering.
Continue the brief
How to Audit Brand Citations Across AI Platforms
Across 297 measured editions, ChatGPT, Gemini, and Google AI Mode agree on the #1 brand only half the time. Single-engine audits are wrong 49.8% of the time.
Did ChatGPT Just Kill Zero-Click for Brand Search?
On May 7 OpenAI made ChatGPT brand mentions clickable. Profound's monitored basket shows referrals up 60% overall and B2B SaaS daily traffic up over 200%.
Has ChatGPT Search Demand Already Peaked?
ChatGPT search demand hit 14,800/mo Jan-Mar 2026 then dropped 33% in April. AI overview and AI search queries are absorbing the demand.
Framework
Learn the CITE framework behind our GEO and AEO work
See how Comprehend, Influence, Track, and Evolve turn AI visibility into an operating system.
Services
Explore our managed GEO services and AEO execution model
Audit, prompt discovery, content execution, and ongoing monitoring tied to AI search outcomes.
Audit
Start with an AI visibility audit before execution
Understand prompt coverage, recommendation gaps, source mix, and where competitors are winning.
