OpenAI today listed the outside firms that will feed conversion data into ChatGPT Ads and report on its results, and added two verification companies for placement-safety pilots, in a post titled "More ways to measure ChatGPT Ads". Among the first results, DV Rockerbox put WeightWatchers' attributed cost per acquisition 15.3% below its blended paid search benchmark, according to OpenAI.

In Short

OpenAI, the company behind ChatGPT, today listed the outside companies that can now help track whether ads shown in the chatbot lead to purchases, sign-ups and app installs. That matters to brands and agencies paying for those ads, because the early figures come from vendors OpenAI chose to feature and arrive without published spend, dates or sample sizes. In practice, sales data can flow in from tools brands already run, reports can credit purchases that follow an ad without a click, and two verification firms are preparing tests of how well ads are kept away from sensitive conversations.

What OpenAI published

The post, filed under Product with the labels Ads and Measurement, opens with an audience claim. According to OpenAI, ChatGPT reaches 1.2 billion people each week, which the company calls the world's largest AI-native consumer platform. The number is about a third higher than the 900 million weekly users OpenAI disclosed on February 27, reached in a little over seven months, and it exceeds the base of more than 1 billion that David Dugan cited around the August 31 revenue milestone.

The rest of the document runs in four parts: connecting campaigns to business outcomes, early results, measuring beyond the click, and brand safety. OpenAI describes its philosophy as supporting both in-house and partner-led measurement, with signals arriving through a pixel and the Conversions API (CAPI), attribution, incrementality and brand measurement. Most of the receiving plumbing is not new. The pixel and CAPI arrived on May 5 with the self-serve Ads Manager, and running both paths together requires deduplication, which the platform's documentation resolves through the pixel ID, event name and event ID. What the post adds is a named list of outside firms sitting on top of that plumbing.

Who plugs in where

Counting each company once, the post names 22 firms, 20 of them in measurement roles (Measured appears in two lists) and two, DoubleVerify and Integral Ad Science, in suitability pilots. OpenAI calls them its first set of third-party partners.

Data connections

According to OpenAI, integrations with Hightouch, Tealium and LiveRamp make it easier to send conversion data from advertisers' existing systems to ChatGPT Ads. Two of the three had surfaced before. LiveRamp connected its CAPI Hub on June 10, Hightouch and Triple Whale appeared in an advertiser email in August, and in late September LiveRamp became a route for customer-list audiences. The post specifies no event schema, match rate or latency for any connection.

Attribution and full-funnel

The attribution list has ten names: AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin. OpenAI says they enable Conversions API, reporting and click attribution solutions across web and app. A second group, described as full-funnel and advanced measurement firms, holds Fospha, Measured and INCRMNTAL.

AppsFlyer and Adjust are the established entries. They went live for app attribution in seven markets in late July, and the post does not say whether the wider roster inherits those market limits.

Experiments

OpenAI says its work on incrementality is "still in its early stages". It is working with Haus, Measured and WorkMagic on geo-based experiments, which split a market into regions, change advertising in some and leave others as a control. The IAB and IAB Europe sorted incrementality methods into four tiers in guidelines dated November 3, 2025, and experimental designs sit at the stronger end of that scale. Whether a study run by a vendor and selected for publication by the platform counts as independent is a separate question the post leaves unaddressed.

Brand studies

Kantar and Cint are in early-stage testing to measure the impact of ChatGPT Ads on brand resonance metrics, a step OpenAI ties to its start of experiments with new placements. No metric, date or placement is given.

The first results, and what they leave out

The post carries a five-card panel of early results. In the copy reviewed, only the first card is legible: DV Rockerbox measured WeightWatchers' attributed CPA on ChatGPT Ads at 15.3% below the company's blended paid search benchmark. The other four cards are not legible and are not reported here.

Attributed figures depend on the credit rules behind them, and the post does not define the benchmark, the spend, the flight dates or the number of conversions. One naming detail also deserves a sentence. DV Rockerbox shares its prefix with DoubleVerify, which is one of the two companies named for the brand suitability pilots. The post does not discuss the relationship.

Two further numbers sit in the body text. According to OpenAI, a geo-lift study by WorkMagic estimated 2.3x more incremental orders for the brand Dose than were captured by last-click attribution. That statement concerns under-crediting by a single attribution rule, and no baseline order count, test period, region list or confidence interval accompanies it.

The third is OpenAI's own analysis. In early data from global campaigns optimized for deeper-funnel outcomes, 52.7% of eligible one-day view-through conversions occurred within an hour of the matched ad impression. A short lag describes when conversions arrive, not whether the ad caused them, because credit and cause are different things in the incrementality framing already described. The count of campaigns, advertisers and conversions behind the percentage is not disclosed, and neither is the split by market, although the platform now spans more than 60 countries.

Beyond the click: windows, optimization and signal quality

OpenAI argues that people often come to ChatGPT to work through a task or decision, so a click may not be the next step, and says it already observes conversions that follow impressions without a qualifying click. The technical answer has two parts, kept deliberately separate.

For reporting, advertisers can select a click-through attribution window that fits their buying cycle and include a one-day window for view-through conversions, meaning purchases that follow an impression with no click. These settings control attribution in reports, and, in the post's words, "they do not change how campaigns optimize."

For delivery, the post says advanced conversion optimization learns from eligible click-through and view-through conversions for the advertiser's selected goal, so that purchases following either path can inform future ad delivery.

That second statement moves a metric across a line. When view-through reporting appeared on August 19 with a fixed one-day window, it was kept out of bidding, billing and optimization. Within weeks, an advertiser email described a model that would optimize through view-through conversions and bill on impressions, and a September email described general availability of impression-billed optimization, configurable attribution windows and an Event Quality Score. The new post describes the optimization behavior but gives no date, billing basis or availability.

The Event Quality Score (EQS) appears in the post as a tool to help advertisers assess the quality of the signals their pixel and CAPI send. The earlier coverage recorded a 1 to 10 grade under Tools and then Conversions in Ads Manager. Signal volume has also been tilted by defaults: automatic advanced matching, which hashes identifiers entered into web forms, was switched on for existing pixels with a ten-day opt-out window in August. The measurement stack also rests on a privacy policy that, as PPC Land reported, disclosed for the first time that OpenAI receives purchase data from advertisers and their partners to measure ad effectiveness. The post's own emphasis sits elsewhere, on keeping conversations private while ads are measured.

Brand safety checks without reading conversations

The final section turns from outcomes to environment. OpenAI says its placement guardrails assess whether a conversation is appropriate for advertising, aiming to keep ads out of "emotionally vulnerable, sensitive, or otherwise unsuitable contexts". Automated review, human oversight and ongoing monitoring identify policy violations. A new control, Negative Phrases, is now available to qualifying advertisers for narrow placement requirements specific to a brand's own policies.

The post gives no qualification criteria, phrase limits or matching rules. Earlier documentation described positive targeting as running on context hints supplied at ad group level, and how a list of excluded phrases interacts with that mechanism is not explained.

OpenAI is also developing pilots with DoubleVerify (DV) and Integral Ad Science (IAS) to build understanding of the ads environment, a brand suitability evaluation. According to OpenAI, the approach allows independent partners to assess how its brand safety standards are applied in a controlled testing environment, without accessing private user conversations. The post supplies two executive statements. Lidiane Jones, chief executive of IAS, says protections need to be "durable, consistent, and built to operate at scale". Mark Zagorski, chief executive of DoubleVerify, describes an environment where "context is defined by the conversation itself rather than a page or video."

Independence is the pitch, and ownership is the complication. DoubleVerify agreed on August 6 to be acquired by Nielsenfor about $2.15 billion, and PPC Land's analysis of that deal asked who checks the checker when IAS sits under private equity and DoubleVerify inside Nielsen. The post mentions neither deal, a pilot start date, a scope or a reporting format, and it does not say whether results will be published.

Why this matters for the marketing community

The sequence matters more than any single number. ChatGPT advertising began on February 9 at a $60 CPM, reached $100 million in annualized revenue within six weeks, and passed a $1 billion run rate on August 31. Prices moved the other way: CPMs were reported as low as $25 by April, a fall PPC Land tied in part to the thin means buyers had of verifying what an impression produced. Revenue arrived first and outside measurement followed, in stages, from the May pixel to today's list.

Buying routes are multiplying in parallel. Amazon DSP buyers gained a US pilot for ChatGPT placements on September 10, and PPC Land noted at the time that neither party had disclosed the attribution infrastructure behind it. A roster of established vendors answers part of that gap, because ChatGPT Ads can now sit inside stacks that advertisers already operate rather than beside them.

Several points remain open in the document itself. The early results are vendor-supplied and arrive without spend, dates or sample sizes. The experiment programme is, by OpenAI's own description, early. Eligibility rules for Negative Phrases, the billing basis of view-through optimization and the start of the DV and IAS pilots are unstated. The next verifiable evidence will be whether any partner publishes methods for a ChatGPT result, and whether the suitability pilots produce a report that anyone outside the pilot can read.

Timeline

Summary

Who: OpenAI, the operator of ChatGPT Ads, with outside firms including Hightouch, Tealium, LiveRamp, AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, Tenjin, Fospha, Measured, INCRMNTAL, Haus, WorkMagic, Kantar and Cint, plus DoubleVerify and Integral Ad Science. The post affects advertisers, agencies and measurement teams buying ChatGPT ads.

What: A post describing conversion-signal connections, attribution and experiment vendors, brand-study testing, a reporting and optimization setup that includes view-through conversions, an Event Quality Score, Negative Phrases and two brand suitability pilots, together with vendor-supplied early results (15.3% lower attributed CPA for WeightWatchers, 2.3x incremental orders for Dose, 52.7% of eligible view-through conversions within an hour).

When: Today. No dates are given for the pilots, the experiments or the brand studies.

Where: Online, within ChatGPT Ads. The 52.7% analysis covers global campaigns, and the post names no individual markets.

Why: According to OpenAI, ChatGPT reaches 1.2 billion people each week, and businesses need ways to understand how advertising in those moments contributes to their goals. The expanded vendor list and safety pilots are meant to give advertisers more ways to evaluate performance while keeping conversations private.