For most of the ChatGPT Ads pilot, mobile app advertisers had a conspicuous hole in the stack: no supported way to prove a click produced an install, let alone what the user did afterward. That changed in July 2026, when OpenAI connected ChatGPT Ads to AppsFlyer and Adjust, bringing app-install and in-app event attribution to seven markets — the US, Canada, Australia, New Zealand, the UK, Japan and South Korea, with Mexico and Brazil flagged as coming next.

So the infrastructure exists. But ChatGPT's ad surface doesn't behave like Meta, Google or TikTok, and configuring it correctly is where the real work sits. Here's what actually needs to be true to run clean, measurable campaigns.

This is part two of a two-part series. Part one covers the strategy — what the channel is, why it matters for mobile, and whether it belongs in your budget. This part is the technical setup.

Where ChatGPT App-Install Inventory Sits

Two surfaces matter for app advertisers today: the sponsored recommendation card that appears below a chat response when a query maps to a product category, and text-style ads in ChatGPT Search alongside web citations. Both render inside the ChatGPT iOS and Android apps for Free and Go tier users — which is the first thing that makes attribution non-trivial.

The Attribution Problem: Why This Isn't Just Another Click Source

Traditional mobile attribution runs on a simple chain: ad click → clickthrough URL → app store redirect → install → MMP match. ChatGPT ads strain that chain in two ways.

The in-app browser environment matters. A click originating inside the ChatGPT app's in-app webview doesn't behave like a click in Safari or Chrome. On iOS, Apple's deterministic flows assume a Safari-rendered surface, so clicks from within the app land in a greyer zone for device-level matching — and post-ATT, probabilistic fingerprinting is legally constrained anyway.

Parameter passing is different. ChatGPT ads don't carry the gclid/fbclid-style click tokens UA teams lean on. Instead, attribution runs through the MMP partner integration and OpenAI's own Conversions API and pixel, which launched in May 2026. Practically, that means the setup is integration-and-postback driven, not click-token driven — which is why getting the MMP config right is the whole game.

iOS vs. Android: The Divergent Reality

Android is materially cleaner. Google Play's Install Referrer can carry campaign-level parameters through the in-app browser to your MMP, and Android doesn't impose the same webview restrictions as iOS. Under a properly configured setup, Android click-to-install attribution is workable — imperfect, but usable for optimization.

iOS leans on SKAN plus modeling. With SKAdNetwork you're working with coarse-grained campaign IDs, short conversion-value windows and no user-level data. The in-app browser context makes deterministic ATT-based attribution even less likely than on Safari. Realistically, iOS attribution for ChatGPT ads today means leaning on SKAN 4.0 postbacks plus modeled incrementality — not deterministic last-click. For iOS-heavy apps (most fintech and travel skew heavily iOS), plan for that from day one rather than discovering it in month two.

Setting Up AppsFlyer for ChatGPT Ads

ChatGPT Ads is now a native AppsFlyer integrated partner, so this is a dashboard setup rather than a custom build. The real flow, per AppsFlyer's published documentation:

Activate the partner. In AppsFlyer, go to Collaborate → Partner Integrations → Marketplace, select your app, search for ChatGPT Ads (OpenAI), and click Integrate. You set this up separately for each platform — your iOS and Android apps are configured independently.

Enter your credentials. The integration needs a Pixel ID and an API key, both generated in the conversions tab of your ChatGPT Ads Manager account. There is no gclid-style click token to map here — the connection is authenticated through the pixel and API key, and attribution flows through server-to-server postbacks. (If you saw an oai_click_id token referenced in earlier write-ups, that's not how the live integration works.)

Configure postbacks and map events. Turn on in-app event postbacks, set your postback window, and map your events to ChatGPT Ads' standard taxonomy. The supported standard events include registration_completed, trial_started, subscription_created, checkout_started, order_created, lead_created, app_opened and contents_viewed, plus a custom fallback. Map deliberately — this is what lets you optimize toward real funnel actions instead of just installs.

Set your attribution link. ChatGPT Ads supports both single-platform links and OneLink for deferred deep linking. Choose your parameters (Campaign Name, Site ID) and set your click-through lookback window — 7 days is AppsFlyer's recommended default, though the fast-convert behavior of high-intent ChatGPT traffic is a reason to consider a shorter window and watch how your conversion curve actually behaves. Your generated link carries openai_int, the integration's ID in AppsFlyer.

Website conversions, if relevant. For web events, the AppsFlyer Web SDK forwards conversions directly to ChatGPT Ads' Conversions API, and OpenAI deduplicates the server-side events against your JavaScript pixel using a shared event ID — so you can run both without inflating conversion counts.

One caveat to plan around: there's currently no cost API for ChatGPT Ads in AppsFlyer, so you won't get spend data pulled automatically into your ROAS reporting yet. You'll need to bring cost in manually until that ships.

Setting Up Adjust for ChatGPT Ads

Adjust brought OpenAI into its partner ecosystem in the same July 2026 wave, so app advertisers can attribute installs and in-app events through Adjust as well. Configuration mirrors the AppsFlyer shape — activate the partner, connect via the pixel/API credentials from your Ads Manager conversions tab, and set up server-to-server callbacks so conversions are matched server-side rather than relying solely on device signals, which partially works around the iOS webview gap.

Where Adjust leans is its privacy-preserving measurement stack. For iOS traffic — where user-level signal will stay structurally incomplete — Adjust's modeled and SKAN-optimization tooling is designed to fill the gaps that postbacks leave, using cohort-level behavioral signals rather than device-level matches. For iOS-heavy apps, treat that modeled layer as a primary measurement signal for this channel, not a nice-to-have.

Because the exact setup fields and SDK requirements are still being iterated across both MMPs, confirm the current minimum SDK version and required entitlements against AppsFlyer's and Adjust's own help centers before you go live — older SDK builds can silently record ChatGPT installs as organic or unknown, which quietly destroys your campaign-level ROAS visibility.

Pre-Campaign Checklist

Before spending a dollar on ChatGPT app-install campaigns:

1. Audit and update your SDK first. No campaign optimization compensates for broken attribution at the SDK layer. Confirm you're on a current AppsFlyer or Adjust SDK build with the ChatGPT partner integration available, before anything else.

2. Register OpenAI as an integrated partner. Activate ChatGPT Ads in your MMP, connect the Pixel ID and API key from your Ads Manager conversions tab, and configure your attribution and postback windows.

3. Set up your SKAN 4.0 conversion schema proactively. For iOS, decide now which in-app events map to your coarse and fine conversion values inside the measurement window. ChatGPT's high-intent profile means early-funnel events (registration, first deposit, first search) tend to fit the window better than late-funnel ones. Calibrate before launch — retroactive schema changes invalidate historical comparisons.

4. Instrument incrementality from day one. Given iOS uncertainty, run a geo holdout or PSA-substitution incrementality test from your first campaign. MMP-reported iOS installs will undercount; incrementality gives you the true causal lift number to defend the budget.

5. Give ChatGPT its own campaign taxonomy. Keep this traffic isolated — don't fold it into "Paid Search" or "Display." The creative, bidding and audience logic are different enough that co-mingling the data corrupts optimization on both sides.

6. Brief creative on conversational formats. These units render inside a response, so copy that reads like a recommendation beats a hard CTA. Short, benefit-forward lines with a clear action noun ("Plan your trip with [App]") tend to outperform "Download Now." It's a different brief than your Meta or UAC team is used to writing.

The Bigger Picture: One Playbook, Many Surfaces

ChatGPT ads are the leading edge of a broader rotation. Perplexity has an ads API, Google's AI Overviews ad unit is scaling, and Meta's AI-assistant surfaces are being commercialized. The common thread: install intent is migrating from keyword-search and interruption-scroll to conversational, problem-statement discovery — and the attribution infrastructure is running a year or more behind the inventory.

Teams that build clean measurement foundations for ChatGPT now are building the playbook they'll reuse across every AI-search ad surface that follows. The MMP ecosystem will standardize and the config will get familiar. But the teams that work out the SKAN schema calibration, the server-to-server callback setup and the incrementality cadence now will carry structural cost and learning-velocity advantages into every channel that comes next.

Don't Wait for the Channel to Mature

The attribution is available today from both AppsFlyer and Adjust. The iOS gap is real but manageable with SKAN 4.0 and modeled measurement; the Android path is cleaner and usable now; the creative requirements are learnable. The cost of waiting isn't only missed installs — it's the attribution data that trains your optimization and informs your channel mix. Entering a high-intent channel without instrumented measurement is the most expensive kind of experimentation there is.

Get your SDK versions right. Configure your MMP partners against their current docs. Build your SKAN schemas. Then run.

Missed the strategy context? Part one covers whether ChatGPT ads belong in your budget at all →

Appvertiser AI helps growth teams automate attribution configuration, creative testing and campaign optimization across emerging UA channels — including ChatGPT Ads. If you're building your measurement stack and want a second set of eyes on the MMP setup, talk to Appvertiser AI.