Measuring a ChatGPT Ads test: the new visual format and attribution partners
By Alan Magliocca · Published November 2026
Quick answer: On October 5, 2026, OpenAI announced a visual ad format shown while Free and Go users generate images in ChatGPT, click attribution through partners including Kochava, Tealium and Triple Whale, early incrementality tests with Haus and WorkMagic, and brand-safety pilots with DoubleVerify and IAS. For a small client, measure a ChatGPT Ads test with the pixel or Conversions API plus tagged URLs, and add a partner only if the client already pays for one.
Key takeaways
- OpenAI's visual ad format appears during image generation on ChatGPT's Free and Go plans, tested with US advertisers from late October 2026.
- Kochava, Tealium and Triple Whale provide reporting and click attribution for ChatGPT Ads.
- Incrementality testing with Haus and WorkMagic is at an early stage, and brand-safety pilots with DoubleVerify and IAS are early too.
- OpenAI's own pixel and Conversions API for ChatGPT Ads launched in May, Marketing Brew reported.
- Most small clients need first-party tracking and a clear decision rule, not a paid measurement stack.
What happened
On October 5, 2026, OpenAI announced a new ChatGPT ad format and measurement tools. OpenAI's site blocks automated reading, so the details here come from Marketing Brew's report by Jasmine Sheena, published October 8 during Advertising Week.
According to Marketing Brew:
- Visual ad format: advertisers can show images of product inspiration or a product in use. It is first being tested while people generate images on ChatGPT's Free and Go plans, starting later in October with a group of US advertisers. OpenAI says the ads are labeled and kept distinct from the images being generated.
- Reporting and click attribution: through Kochava, Tealium and Triple Whale.
- Brand lift: testing with Kantar and Cint.
- Incrementality: tools are in development, with early tests alongside Haus and WorkMagic.
- Brand suitability: early pilots with DoubleVerify and Integral Ad Science (IAS), and wider availability of Negative Phrases, which keeps ads away from queries containing terms a brand considers unsuitable.
- Earlier tools: OpenAI's own Conversions API and pixel for ChatGPT Ads launched in May.
Marketing Brew's report did not include advertiser counts or minimum spends.
What it means for agency owners
Clients ask two questions before they commit budget to a new channel: did it work, and is it safe for my brand? ChatGPT Ads now has answers to both that look like what clients know from Google and Meta.
For a small local client, most of the new stack is optional. Kochava, Tealium and Triple Whale are paid tools that larger advertisers already use. If your client doesn't use one, adding it for a first test usually costs more than the test is worth. The pixel or Conversions API, plus tagged URLs, covers what a small test needs.
The visual format is a different kind of placement. It shows while people make images, not while they ask for advice. That suits products people picture, such as home decor, fashion or renovation, more than services people research.
Incrementality is the honest question in any new channel: would these sales have happened anyway? OpenAI's tools for that are early. Until they mature, a simple holdout or a before-and-after comparison is the best a small client can do.
What to do this week
- Set up first-party tracking before any spend: the ChatGPT Ads pixel or Conversions API if Ads Manager offers it in the client's market, and UTM tags on every ad URL.
- Decide the one result that counts, such as a booked call or a purchase, and make sure it is recorded in the client's CRM or store.
- Ask whether the client already uses Kochava, Tealium or Triple Whale. If yes, connect it. If no, skip it for now.
- Set Negative Phrases for topics the client never wants to appear next to.
- Write the decision rule down: budget, dates and the cost per result that would justify continuing.
- For product clients, ask about the visual format only if they sell something people picture while making images. It started with a US test group.
How AIVX Labs helps
AIVX Labs doesn't connect to ChatGPT Ads or track conversions, so measurement setup happens in Ads Manager and the client's own tools. Where it helps is the reporting step. The monthly client report puts the results, and what you did, in front of the client in plain terms. The monthly client report demo shows RICO building one, and we cover the template in the monthly what-we-did-for-you client report.
Which measurement layer does a small client need?
Think of measurement as three layers, and add one only when the one below it works.
- Layer 1: first-party tracking. Pixel or Conversions API, UTM tags and the client's own records. Every test needs this.
- Layer 2: an attribution partner. Kochava, Tealium or Triple Whale. Worth it when the client already uses one, or spends enough across channels that comparing them is the main question.
- Layer 3: incrementality. A test that shows whether ChatGPT Ads caused sales the client wouldn't otherwise have made. OpenAI's partner tests are early, so this is mainly for larger budgets for now.
Most local clients stop at layer 1, and that is fine.
A worked example: measuring a $1,500 test
Picture a kitchen showroom that agrees a $1,500 ChatGPT Ads test over 30 days, about $50 a day.
- Tracking: the pixel fires on the "book a design visit" confirmation page, and every ad URL carries utm_source=chatgpt.
- Records: the showroom's booking calendar tags each visit with its source.
- Decision rule: if the cost per booked design visit is within 20% of their Google Search cost per visit, extend for another 30 days.
- Holdout: if the showroom runs no other new marketing in that period, compare total bookings with the same 30 days before the test.
- Review: at day 30, look at bookings, not clicks.
If the cost per visit is far above Google's, stop and keep the landing page for other channels. The rule protects both of you, whatever the answer.
For the setup side of a first test, see ChatGPT Ads in 89 countries: a first-test plan. If the client uses HubSpot, the CRM link changes how leads are counted; see what a Sponsored Agents lead looks like.
The same measurement thinking applies to event funnels. BrightStage's guide on how to measure webinar ROI works through it for webinars.
Sources
Related
Written by Alan Magliocca, founder of AIVX Labs and BrightStage AI (Podavinci LLC). Written with AI assistance and checked against the linked sources.
Read our featured article on LinkedIn
Get AIVX Labs verified for $1: RICO, your AI operator, with 7 days of full access.
Get AIVX Labs Verified for $1Follow along:LinkedInYouTubeTikTokXFacebookInstagramSkool community
Frequently asked questions
Who sees the visual ChatGPT ads?
Marketing Brew reported they are first tested while people generate images on ChatGPT's Free and Go plans, starting in late October 2026 with a group of US advertisers.
Do I need Triple Whale or Kochava to run ChatGPT Ads?
No. They are optional partners for reporting and click attribution. OpenAI's own pixel and Conversions API, plus UTM tags, are enough for a small test.
Can I prove ChatGPT Ads caused the sales?
Not precisely yet for a small client. OpenAI's incrementality tests with Haus and WorkMagic are at an early stage, so use a clear decision rule and, where possible, a before-and-after comparison.
