ChatGPT Ads has changed since the early self-serve rollout. OpenAI's September 16, 2026 announcement added AI-assisted campaign tools and integrations, alongside a limited test of Sponsored Agents. Advertisers should separate those announcements from the features actually enabled in their accounts.
Campaign lab
New ad features matter only when the measurement path is ready.
The interesting part of ChatGPT Ads is not a familiar campaign setting with a new label. It is the placement beside a live decision. That makes context, creative continuity, landing-page clarity, and conversion measurement more important than copying an old paid-search account into a new interface.
This update was reviewed on September 28, 2026. For the basic campaign workflow, use our ChatGPT Ads preparation guide.
What changed in September 2026?
In its September announcement, OpenAI described Sponsored Agents being tested with select US advertisers, campaign management through the Ads Manager plugin, AI creative assistance, opt-in text customization, and integrations with HubSpot and Shopify.
A Sponsored Agent lets someone choose a separate, labeled conversation with a business after an ad. It does not turn the original ChatGPT answer into the advertiser's message. The test should not be presented as an option every advertiser can activate.
What should a team verify before using a new feature?
| Area | Verification question | Evidence to save |
|---|---|---|
| Access | Is this feature enabled for our account and market? | Account screen and current documentation |
| Creative assistance | Who approves generated claims and images? | Approved assets and revision record |
| Text customization | Which messages may change, and which claims require exact wording? | Settings and approved claim boundaries |
| CRM or commerce integration | Do campaign IDs and conversion events reconcile? | Test event and reporting comparison |
| Sponsored Agent test | Are we eligible, and who maintains its product information? | Access confirmation and content owner |
A product announcement is a reason to investigate. It is not a substitute for confirming account access or testing the customer's path.
Choose the objective before the creative
The current advertiser overview describes impression, click, and conversion-oriented buying. For conversion-optimized campaigns using click billing, the charge is for valid clicks, not completed conversions.
That difference belongs in the campaign brief. “Optimize toward a lead” and “pay only for a lead” are different promises. Choose the objective available in the account, document its billing basis, and check the configured event before setting the budget.
Keep the offer consistent from ad to page
Suppose a service ad offers a technical audit. The landing page should explain what is examined, what the buyer receives, what it costs or how a quote is determined, and how to book. Sending that person to a broad homepage makes the promise harder to evaluate.
For a catalog campaign, check the product title, image, price, availability, and destination together. A correct ad with an outdated landing page still creates a poor experience. Treat feed and page maintenance as part of campaign operations.
Review AI-assisted creative like any other advertisement
AI assistance can speed up a draft; it cannot approve your claims. Check that a generated image represents the offer accurately and that the copy does not invent certifications, discounts, customer results, or availability.
Keep a short approved-claims sheet alongside the campaign. Give each claim a source and an owner. If a claim cannot be supported, remove it before the ad is used. This makes review faster as the number of creative variations grows.
Measure the complete path
Keep campaign, ad group, and creative names consistent with the analytics records. Test the landing page and conversion event on a real browser before launch. Compare platform reporting with your own records, allowing for differences in attribution windows, consent, timing, and event definitions.
For a lead-generation campaign, distinguish a submitted form from a qualified inquiry and a won customer. A low cost per form is not useful if the people submitting it cannot use the service.
For an initial test, write down:
- The audience problem and offer.
- The primary conversion and its definition.
- The budget and review schedule.
- The conditions for changing, pausing, or expanding the test.
- The person responsible for approving creative and investigating tracking errors.
How does this affect GEO and organic AI visibility?
Paid delivery and organic citations answer different questions. Keep paid performance in the campaign report and organic mentions in a separate AI visibility report. Do not count a sponsored impression as an earned recommendation.
The shared work is information quality. Accurate product facts, clear explanations, and a useful destination help a reader evaluate the business wherever they find it. Our conversational advertising overview compares that preparation across platforms.
Sources and further reading
September announcementcurrent advertiser overviewCampaign path
Walk one ad test from decision moment to measured conversion.
Use the stages to keep the message, landing page, and conversion event connected.
Define the decision moment
Describe the situation in which the offer is genuinely useful. Keep one decision moment per test.
Quick answers
Questions people usually ask next.
Do ChatGPT Ads influence organic answers?
No. Sponsored placement and organic answer visibility should be planned and measured as separate systems.
What should be ready before the first campaign?
A focused objective, tested conversion event, narrow context themes, appropriate creative, a matching landing page, and a documented measurement plan.
Should every advertiser use every new Ads Manager feature?
No. Product feeds, audience tools, and advanced bidding are useful only when they fit the offer, data quality, privacy requirements, and campaign objective.
Put the research to work
