OpenAI has connected ChatGPT Ads to the two largest mobile measurement partners, giving app advertisers install and in-app event attribution for the first time since the pilot began in February 2026. AdTechRadar reported the integrations on July 24, 2026, citing a product update email sent to advertisers in the ChatGPT Ads Manager beta.
The addition closes one of the more conspicuous holes in OpenAI's advertising stack. Until now, an app marketer buying placements inside ChatGPT had no supported route to prove that a click produced a download, let alone what the downloader did afterwards. Both AppsFlyer and Adjust have now published setup documentation on their own help centres, and the AppsFlyer article carries a last-updated stamp of July 9, 2026 at 2:24 PM, which places the technical launch at least two weeks ahead of the trade press report.
According to OpenAI, the mobile measurement work was one of eight items in a single weekly product note. "This week's product updates focus on helping advertisers optimize toward conversion goals, improve conversion measurement, pace budgets more effectively, and scale campaign operations," the company told advertisers in the email. On the measurement partners specifically, the note read: "We recently launched integrations with AppsFlyer and Adjust to measure and attribute app installs and in-app events from ChatGPT Ads."
What Adjust receives, and how
Adjust classifies ChatGPT Ads as an ad network within its system rather than as a self-attributing network. That distinction is not cosmetic. It means the integration runs on automated callbacks: Adjust performs the attribution and pushes the resulting data outward, rather than accepting a partner's own claimed conversions and reconciling them afterwards.
Setting the connection up requires Admin, Editor, or Custom Editor permissions in Adjust, plus SDK version 4.0.0 or later if an advertiser intends to map SDK parameters across to ChatGPT Ads parameters. The supported development environments listed in the documentation span iOS, Android, Windows, Unity, React Native, Flutter, Adobe Air, Cordova, Marmalade, Xamarin, Cocos2d-x, Titanium and Corona - a breadth that suggests the network entry was built on Adjust's standard partner template rather than as a bespoke case.
Five credentials are needed to activate the link: API KEY, PIXEL ID, ACTION SOURCE, INSTALL EVENT NAME, and an optional SOURCE URL. The pixel identifier is the same one used for web measurement, which indicates that OpenAI is running app and site conversions through a shared identity spine rather than two parallel systems. Advertisers are directed to their ChatGPT Ads representative to obtain the values.
Once data sharing is enabled, Adjust sends install and event data by default. Beyond that baseline, the advertiser chooses. Under the data sharing settings, a campaign can be configured to forward activity from all attribution sources or only activity attributed to ChatGPT Ads itself - the difference between giving OpenAI a full picture of app behaviour and giving it only the slice it earned. Multi-platform apps can apply platform-specific settings or a single global rule.
Event mapping follows the same logic. Data for unmapped events is not shared at all, so an advertiser who does nothing shares nothing beyond installs. Individual Adjust events can be mapped to named partner events, or, where the network supports it, all events can be mapped at once. Partner parameters, available from SDK 4.0 onward, carry custom data points from the app itself.
Three placeholders are mandatory on every callback. Adjust must send {currency}, the original ISO 4217 currency code as transmitted by the SDK; {revenue_float}, revenue in whole currency units; and {created_at_milli}, an activity timestamp carrying millisecond precision. The revenue requirement is the more telling of the three. A network that insists on receiving monetary value alongside every event is building toward value-based optimisation, not merely counting conversions.
Completing the setup generates a click URL, an impression URL, a QR code and a link token. The impression URL matters for view-through measurement, a signal type that has not previously featured in public descriptions of how ChatGPT ads are credited.
AppsFlyer covers apps and websites in one integration
The AppsFlyer implementation is broader in scope. It handles two distinct measurement jobs under a single partner entry, and each has to be configured separately for every app and every site.
On mobile, AppsFlyer attributes installs and re-engagements to ChatGPT Ads campaigns and sends in-app event postbacks. On web, it forwards events captured by the AppsFlyer Web SDK - page views, content views and purchases among them - to ChatGPT Ads' Conversions API, either alongside the ChatGPT Ads JavaScript pixel or instead of it.
The parallel-running case is where the engineering detail sits. According to AppsFlyer, when an advertiser fires both the pixel and server-side CAPI events, ChatGPT Ads deduplicates the two streams against each other using a shared event ID, so conversion counts are not inflated. That is the same deduplication pattern Meta established for its own Conversions API, and its presence here is a signal that OpenAI has adopted the conventions of the incumbent platforms rather than inventing its own.
Configuration runs through Collaborate, then Partner Integrations, then the Marketplace tab, with five configuration surfaces per integration: Integration, Attribution link, Cost, Ad revenue, and Permissions. A Pixel ID and API key from the advertiser's ChatGPT Ads Manager account are required for each app and site. Android and iOS builds of the same app must be set up separately.
The presence of a Cost tab is worth pausing on. Cost ingestion means AppsFlyer expects to pull spend data back from OpenAI, which is what makes return-on-ad-spend calculation possible inside the MMP dashboard. Without it, an advertiser can count installs but cannot price them.
Seven live markets, two pending
AppsFlyer's traits and limitations table sets out the geographic boundary of the integration with unusual precision. ChatGPT Ads campaigns are supported in the United States, Canada, Australia, New Zealand, the United Kingdom, Japan and South Korea. Mexico and Brazil are described as coming soon.
That list maps exactly onto the pilot's activation history. Advertising went live in ChatGPT for free and Go tier users in the United States on February 9, 2026, with Canada, Australia and New Zealand added in March alongside disclosure of $100 million in annualised revenue inside six weeks. The United Kingdom followed on June 6, and Japan and South Korea on June 22, the same day the self-service Ads Manager beta opened to British advertisers.
Mexico and Brazil were named in the same May 7 expansion plan that produced the other five markets. Their absence from the supported list, and their presence on the pending one, is the clearest public indicator available that neither has yet been switched on.
The rest of the July update
The measurement partnerships arrived bundled with seven other changes, several of which alter how money is spent rather than how outcomes are counted.
Conversion-optimized campaigns are now created by selecting the Conversions objective during campaign setup, producing what OpenAI describes as conversion-optimized cost-per-click, or oCPC. Delivery is steered toward clicks judged more likely to produce the advertiser's conversion event, while billing remains tied to valid clicks. The mechanism preserves the CPC billing model while introducing conversion-weighted delivery, an arrangement that shifts prediction risk onto the platform. Conversion-optimized campaigns first rolled out on June 5 to advertisers who had a conversion event flowing through the pixel or Conversions API by June 1.
A daily budget change was flagged as arriving the following week. Fixed daily limits become average daily budgets, defined as the average an advertiser wants to spend per day across seven days. Total budgets do not change and no advertiser action is required, but daily spend will vary while remaining within both the daily and the seven-day ceilings. Separately, daily budgets now pace automatically through the day. Daily budgets themselves were only introduced on May 22, alongside US geo-targeting by state, designated market area and ZIP code; the shift to averaging arrives roughly two months later.
Geo exclusion now permits specific locations to be removed from campaign targeting, the inverse of the inclusion controls added in May.
Automatic advanced matching, abbreviated AAM in the documentation, improves website conversion measurement by matching additional conversions to ads using hashed customer information. It is enabled through Tools, then Conversions, then Data Source, by selecting the option to enable for all web data sources. The hashing approach mirrors the identifier handling in ChatGPT's Custom Audiences feature, which requires a minimum of 25,000 matched users before an audience can be used.
A Bulk API allows campaigns, ad groups and ads to be created and updated asynchronously and at scale through the Ads API. Bulk operations are a prerequisite for agency and in-house teams running large account structures, and their absence has been a practical constraint on how much budget could be managed inside the platform.
Finally, a refreshed product card format for product feed campaigns is beginning to roll out, adding price and star ratings to the creative. The sample card shown in OpenAI's email displays a cordless vacuum at $249.99 with a 4.6-star rating and a merchant name beneath. Product feed support was added to Ads Manager in early June, reusing the structured files retailers already submit to Google Shopping.
Why this matters for the marketing community
Attribution has been the persistent complaint about ChatGPT advertising since the channel opened. CPMs fell from $60 at launch to as low as $25 by April partly because buyers had limited means of demonstrating what the impressions produced. A Conversions API and pixel arrived on May 5 with the self-serve Ads Manager, and LiveRamp connected its CAPI Hub on June 10 to route server-side conversion signals. Each step addressed web. None addressed apps.
The gap mattered commercially. Mobile app budgets are among the most measurement-dependent in digital advertising, because the install is a discrete, verifiable event and because the downstream value of a user can be modelled with unusual precision. An MMP sits as a neutral layer between advertiser and platform, applying its own deduplication logic and attribution windows to decide which touchpoint gets credit. Without one, an app advertiser buying ChatGPT placements was relying entirely on OpenAI's own reporting of OpenAI's own performance.
The choice of partners is not arbitrary. AppsFlyer is generally regarded as the largest MMP in operation; Adjust, owned by AppLovin following a reported $1 billion acquisition in 2021, is usually placed second. Together they cover the majority of the app advertising market's measurement infrastructure, and both hold badged status with Meta, Google and TikTok. An advertiser already running app campaigns across those channels can now compare ChatGPT performance inside the same dashboard, using the same attribution rules, without building anything new.
AppsFlyer's own position has shifted in parallel. Google, Meta, Moloco and Unity invested in the company at a $2.7 billion valuation on June 22, 2026, in a deal explicitly structured to preserve measurement neutrality. Adding OpenAI as an integrated network one month later extends that neutrality claim to a platform none of the four investors control.
There is a counterweight. Adthena recorded zero paid ChatGPT placements across 169,560 UK result scrapes during June 2026, even as the same firm measured Google AI Overviews on 23 percent of tracked UK queries. Measurement infrastructure is being built faster than demand is arriving in some markets, and the divergence between competing third-party datasets remains unresolved. Adthena separately counted 7,378 distinct ChatGPT advertisers in the week of July 13 to 20, 2026, with the United States accounting for 60.1 percent of them.
What the MMP integrations change is the class of advertiser that can plausibly test the channel. A performance team with an app, a bid target and an existing AppsFlyer or Adjust contract now faces a setup cost measured in configuration screens rather than in custom engineering. Whether the conversational context that OpenAI sells converts to installs at a rate that justifies its pricing is a separate question, and one that only the new attribution data will answer.
Timeline
- February 9, 2026 - ChatGPT advertising launches for logged-in Free and Go tier users in the United States
- March 26, 2026 - Pilot crosses $100 million in annualised revenue; Canada, Australia and New Zealand confirmed
- April 17, 2026 - ChatGPT ad CPMs fall from $60 at launch to as low as $25
- May 5, 2026 - Self-serve Ads Manager beta opens to all US businesses with CPC bidding, Conversions API and pixel
- May 14, 2026 - Custom audience targeting appears in a gated rollout
- May 22, 2026 - Daily budgets, US geo-targeting and dynamic call-to-action tests added
- June 2, 2026 - Product feed support confirmed for Ads Manager
- June 5, 2026 - Conversion-optimized campaigns begin rolling out
- June 6, 2026 - Advertising activates in the United Kingdom
- June 10, 2026 - LiveRamp connects its CAPI Hub to ChatGPT Ads
- June 22, 2026 - Ads go live in Japan and South Korea; self-serve beta opens to the UK
- June 22, 2026 - Google, Meta, Moloco and Unity invest in AppsFlyer at a $2.7 billion valuation
- July 9, 2026 - AppsFlyer publishes its ChatGPT Ads integration documentation, last updated at 2:24 PM
- July 13 to 20, 2026 - Adthena observes 7,378 distinct ChatGPT advertisers across its tracked markets
- July 24, 2026 - AdTechRadar reports the AppsFlyer and Adjust integrations, citing OpenAI's advertiser email covering oCPC campaigns, average daily budgets, spend pacing, geo exclusion, automatic advanced matching, a Bulk API and refreshed product cards
Related PPC Land coverage
- OpenAI opens ChatGPT Ads Manager to all US businesses with CPC bidding - Documents the May 5 arrival of the Conversions API and pixel that the AppsFlyer web integration now feeds.
- OpenAI's ChatGPT ads are getting conversion optimization - here's what changes - Explains the conversion objective and the June 1 eligibility cutoff that preceded the oCPC rollout.
- ChatGPT Ads Manager gets daily budgets, geo-targeting, and dynamic CTAs - Covers the May introduction of the daily budgets now being converted to seven-day averages.
- ChatGPT Ads Manager now supports product feeds after checkout is killed - Sets out the product feed mechanics behind the refreshed price-and-ratings card format.
- LiveRamp CAPI Hub now connects ChatGPT ads - here's what it measures - The previous third-party measurement connection into the platform, covering web rather than app.
- Google, Meta, Moloco, and Unity buy into AppsFlyer to keep measurement neutral - Details the June investment and the neutrality argument that frames AppsFlyer's role here.
- The four MMP selection mistakes costing app marketers their data - Background on what mobile measurement partners do and how attribution windows differ between them.
- ChatGPT ads go live in Japan and South Korea - and UK gets self-serve - Establishes the seven-market footprint that AppsFlyer's supported-countries table now mirrors.
- OpenAI blocks ChatGPT ad audiences under 25,000 matched users - Describes the hashed-identifier handling that automatic advanced matching extends to web conversions.
- UK advertisers face zero ChatGPT ad share, Adthena data show - The counter-signal on demand-side adoption in one of the seven supported markets.
- Adthena counts 7,378 ChatGPT advertisers as placements grow 97x in a quarter - Current third-party count of advertiser participation across the tracked markets.
- 58.6% of marketers underinvest in channels AI can't measure, AppsFlyer finds - Research tying measurement gaps to budget allocation, published by one of the two integration partners.
Summary
Who: OpenAI, operator of the ChatGPT Ads platform, together with mobile measurement partners AppsFlyer and Adjust, the latter owned by AppLovin. The changes affect app advertisers, performance agencies and retailers running product feed campaigns inside ChatGPT.
What: ChatGPT Ads has been added as an integrated partner in both AppsFlyer and Adjust, enabling install attribution, re-engagement attribution and in-app event postbacks for ChatGPT campaigns. AppsFlyer additionally forwards web events from its Web SDK into ChatGPT Ads' Conversions API, with server-side events deduplicated against JavaScript pixel events via a shared event ID. Adjust operates the connection as an ad network with automated callbacks, requiring API key, pixel ID, action source, install event name and optional source URL, and forwarding currency, revenue and millisecond timestamp parameters. The integrations shipped alongside conversion-optimized cost-per-click campaigns, a shift from fixed to average daily budgets, intraday spend pacing, geo exclusion, automatic advanced matching, a Bulk API and a refreshed product card format carrying price and star ratings.
When: AppsFlyer's integration documentation carries a last-updated timestamp of July 9, 2026. AdTechRadar reported the integrations and the accompanying product update on July 24, 2026. The average daily budget change was described as arriving the following week.
Where: ChatGPT Ads campaigns are supported in the United States, Canada, Australia, New Zealand, the United Kingdom, Japan and South Korea, with Mexico and Brazil listed as coming soon. Automatic advanced matching and the Bulk API operate through the ChatGPT Ads Manager and Ads API respectively.
Why: App install measurement had no supported path inside ChatGPT Ads before these integrations, leaving mobile advertisers dependent on OpenAI's first-party reporting. Connecting the two largest MMPs places ChatGPT campaign data in the same dashboards, under the same attribution rules, that app marketers already apply to Meta, Google and TikTok, which lowers the operational cost of testing the channel and makes cross-channel comparison possible for the first time.
Discussion