ChatGPT Ads measurement partners and a visual ad format test
On 5 October 2026 OpenAI expanded ChatGPT Ads measurement and announced a visual ad format test. Conversion data can now flow in through Hightouch, Tealium and LiveRamp; ten attribution partners support Conversions API, reporting and click attribution; three partners are exploring geo-based incrementality. The visual format will be tested from late October in the US with an initial group of advertisers, shown during image generation to users on the Free and Go plans. Negative Phrases are available to qualifying advertisers, and brand-suitability evaluations with DoubleVerify and IAS are pilots.
| Platform | ChatGPT Ads |
|---|---|
| Feature | ChatGPT Ads measurement integrations |
| Announced | Oct 5, 2026 |
| Status | Rolling out |
| Where available | Measurement integrations and attribution partners: announced as supported now. Visual ad format: testing from late October 2026 in the US with an initial group of advertisers, for users on Free and Go plans. Negative Phrases: available to qualifying advertisers. Brand-suitability evaluations with DoubleVerify and IAS: pilots. |
| Who is affected | Advertisers running ChatGPT Ads who report CPA or ROAS, teams evaluating the channel, and brands with placement requirements. |
| Action needed | Optional. Connect conversion data through your existing provider if it is on OpenAI’s list, agree a measurement basis before comparing with Search or social, and treat the visual format as a separate test. |
| Sources | OpenAI: Building advertising for the way people use AI |
What changed
ChatGPT Ads are labelled ads shown to ChatGPT’s Free and Go users, selected from the context of the conversation; OpenAI says they do not influence ChatGPT’s answers and that advertisers never see the conversation. Until this update, measurement leaned on OpenAI’s own pixel and Conversions API. The announcement adds three layers:
- Conversion data in: integrations with Hightouch, Tealium and LiveRamp so advertisers can send conversion data from systems they already run.
- Attribution partners: Conversions API, reporting and click attribution through AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin, plus full-funnel and advanced measurement with Fospha, Measured and INCRMNTAL.
- Incrementality: geo-based experiments with Haus, Measured and WorkMagic, which OpenAI describes as early-stage.
OpenAI cites three partner-reported results: DV Rockerbox put WeightWatchers’ attributed cost per acquisition on ChatGPT Ads 15.3% below its blended paid-search benchmark; WorkMagic reported statistically significant lift for the wellness brand Dose with 67% of incremental purchases from net-new customers; and Triple Whale reported that 93% of Portland Leather’s visitors from ChatGPT Ads were new. These are partner figures for individual advertisers, not platform averages.
The visual ad format places images, product inspiration, product usage or the experiences a product enables, during image generation in ChatGPT. OpenAI says ads will be clearly labelled and separate from the generated image, that testing begins later in October in the US with an initial group of advertisers, and that it applies to Free and Go plan users. There is no self-serve route to it yet.
On brand safety, OpenAI describes placement guardrails that keep ads out of emotionally vulnerable, sensitive or otherwise unsuitable conversations, backed by automated review and human oversight. Negative Phrases, a per-brand exclusion control, are “now available to qualifying advertisers.” Brand-suitability evaluation pilots with DoubleVerify and Integral Ad Science will let those firms assess how OpenAI’s standards are applied in a controlled environment without access to private conversations.
Why it matters
The partner list is the real news. Most performance teams already run one of AppsFlyer, Triple Whale, Northbeam or Adjust, and the question that kept ChatGPT Ads in the “experimental budget” column was whether it could be measured on the same basis as everything else. Now it can be, at least for click attribution and Conversions API. That moves the channel from a curiosity to a line in the same report as Meta and Google, with the same caveats about attribution models.
Read the case figures carefully. A 15.3% lower attributed CPA than paid search, 93% new visitors, 67% net-new incremental purchases: each is one advertiser, one partner, one model. The high new-customer shares are plausible, because people asking ChatGPT for recommendations are often early in a decision, and that is the strongest argument for the channel. But they are also exactly the kind of number that gets copied into a forecast. Our view: run ChatGPT Ads through the attribution tool you already use so the comparison is like-for-like, hold a geo test for later once incrementality partners have more to show, and keep the visual format out of any plan until it reaches your account. The brand-safety pilots are worth watching for regulated categories; a DoubleVerify or IAS report will carry more weight with a compliance team than OpenAI’s own description.
What to do
- Check whether your attribution or CDP provider is on OpenAI’s list, then ask them which ChatGPT Ads connection is live for your account, which events it passes and how it deduplicates against the OpenAI pixel.
- In ChatGPT Ads Manager, confirm the Conversions API or pixel events match the conversion definition and attribution window you use for Google and Meta before putting the channel in a shared report.
- Reconcile a two-week sample of attributed conversions against your own orders or qualified leads, and note the share of new customers in your own data rather than the partner’s.
- If your brand has placement requirements, ask your OpenAI contact whether your account qualifies for Negative Phrases and what the qualification is.
- Keep any visual-format invitation as a separate campaign with its own budget so a format change does not become an untracked variable.
Still unknown
OpenAI has not published the qualification criteria for Negative Phrases, the size or selection of the visual-format test group, or a date beyond “later this month.” The incrementality work is described as early; no methodology or results are published. The three case figures come from partners and have no stated periods or sample sizes. We have not run ChatGPT Ads through these integrations ourselves.
Frequently asked questions
Which measurement partners does ChatGPT Ads support?
OpenAI’s 5 October 2026 announcement names Hightouch, Tealium and LiveRamp for sending conversion data; AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin for Conversions API, reporting and click attribution; Fospha, Measured and INCRMNTAL for advanced measurement; and Haus, Measured and WorkMagic for geo-based incrementality experiments.
What is the ChatGPT visual ad format?
An image-based ad shown during image generation in ChatGPT, clearly labelled and separate from the image being created. OpenAI said testing would begin later in October 2026 in the US with an initial group of advertisers, for users on the Free and Go plans.
What are Negative Phrases in ChatGPT Ads?
A brand-level control that keeps ads away from conversations matching phrases the advertiser specifies, on top of OpenAI’s own placement guardrails. OpenAI says they are available to qualifying advertisers; it has not published the qualification criteria.
Which ChatGPT users see ads?
Users on the Free and Go plans. OpenAI says Plus, Pro and Enterprise subscriptions remain ad-free, and that advertising does not influence ChatGPT’s answers.
Official sources
- OpenAI: Building advertising for the way people use AI ↗ · Oct 5, 2026
Related coverage
- ChatGPT Ads availability by country: self-service list, 2026
- ChatGPT Ads in Shopify and HubSpot, plus AI copy in Ads Manager
Update history
- — Rewritten under the new editorial standard with the full partner list, the visual-format test conditions, OpenAI’s partner case figures and their sources, and a practitioner view on measurement.
- — Clarified the feed and metadata description: the visual test is planned, with access unconfirmed.
- — Correct the headline: October visual-format testing is planned, not confirmed started. Keep the measurement expansion and account-access caveats.
- — Initial manual coverage under the advertiser-impact standard; original announcement date preserved.
Suggested checks are editorial guidance, not platform requirements. Every claim links to its official source. Report a correction →