Preview|APP & U 3Q26: Data Drama

We recently spoke with several advertising channels, leading gaming and e-commerce clients, and multiple publishers to follow up on the following key topics.

• The AppLovin vs. Unity litigation
• Publisher feedback on the impact on AppLovin’s acquisition costs
• Changes in AppLovin vs. Unity competition since 2Q26
• AppLovin’s bids compared with other platforms
• AppLovin quarterly survey
• AppLovin e-commerce progress
• Impact of Meta Audience Network on AppLovin
• CloudX progress
• Unity advertiser feedback
• AppLovin quarterly view and model update
• Unity quarterly view

I. The AppLovin vs. Unity litigation

In last quarter’s AppLovin and Unity Previews, we specifically noted that AppLovin’s IAA D28 business faced intense competition from Unity in 2Q26. In the Unity Preview, we highlighted that Unity’s IAA D28 business achieved in just one quarter nearly the growth that its IAP D28 business took several quarters to achieve, which directly hit AppLovin’s core advertising business.

This view was confirmed by the legal dispute between AppLovin and Unity that began in August. A careful analysis of the dispute’s timeline reveals many meaningful clues.

1. AppLovin’s complaint

AppLovin and Unity signed the “AppLovin Header Bidding Agreement” on October 13, 2021, under which Unity connected to AppLovin’s mediation platform MAX as a bidder. Section 2.4(c) of the agreement provides that bid request data may be used “solely to bid on Inventory made available via the Mediation Platform”.

AppLovin alleges that, since at least May 2025, Unity used the Unity Ad Quality SDK to scrape data on ads that AppLovin won and displayed, including creatives, click-through landing pages, advertiser and campaign IDs, device and user IDs, revenue per impression, MAX bidding and waterfall information, and even the coordinates of the most recent screen tap.

AppLovin alleges that Unity’s scraping method was to “hook into AppLovin callbacks and inject Unity code into its ads”, an active behavior rather than passive observation.

AppLovin suspects Unity used this data to train Vector, its competing bidding model, but this is speculation; its public filings contain no direct evidence.

2. Timeline

October 13, 2021: AppLovin and Unity signed the “AppLovin Header Bidding Agreement”.

May 28, 2025: Unity LevelPlay 8.9.0 began bundling Ad Quality; AppLovin says Unity’s data collection began at least from this point.

August 5, 2026: AppLovin reported 2Q26 results. 2Q26 revenue was about 1% below the high end of guidance, short of its usual beat margin, but it gave 3Q26 revenue guidance implying q/q acceleration.

August 21, 2026: AppLovin sent a cease-and-desist letter, requiring Unity to respond within 14 days.

September 4, 2026: Unity replied with a refusal: “Ad Quality does not collect any data from AppLovin”

September 8, 2026 and September 14, 2026: AppLovin issued 72-hour and 48-hour ultimatums in succession, with additional demands.

September 11, 2026: Unity made its first concession, agreeing to notify publishers within 5 business days to disable Ad Quality on MAX, and to release an SDK without Ad Quality within 30 days.

September 16, 2026: Unity made its second concession, agreeing to remotely switch off collection in MAX bidding (including ironSource Ads) and to delete collected data under its normal retention policy; it refused to retrain or take down its model; all conditional on AppLovin confirming in writing that the dispute was settled.

September 27, 2026: AppLovin filed for arbitration with JAMS

October 2, 2026: The court denied the TRO and the request for a preliminary injunction hearing

3. Unity’s stance on the matter

In its September 4 letter, Unity acknowledged the data collection, which it said had taken place “since well before the parties entered into those agreements”.

Unity acknowledged that the collection targeted ads won in MAX bidding, with the data coming from publisher apps and user devices, and with authorization (it denied the data came from AppLovin, but did not deny collecting data on MAX ads).

Unity argues that AppLovin knew about it long ago.

Unity can control the data collection remotely. It can remotely switch off collection in MAX bidding, including Ad Quality deployed within ironSource Ads. Unity does not agree to retrain the Vector model.

Unity alleges that AppLovin’s Ad Review is far more invasive, is enabled by default for all MAX publishers, and also collects impression data for Unity Ads and others; it also argues that MAX gives the AppLovin DSP additional support.

4. Can other DSPs replicate Unity’s behavior?

Back in 2020, Mintegral used the Mintegral SDK to intercept competitors’ ad click data in thousands of apps. After this was exposed, platforms delisted Mintegral, and Mintegral subsequently stopped the practice. AppLovin cited this incident in its filings, saying Ad Quality is of the same nature but more serious.

Most DSPs do not have this capability. However, platforms that have both DSP and mediation capabilities can obtain data through ad quality tools like Ad Review and Ad Quality. The main function of these ad quality tools is to record displayed ad creatives and block ads that are non-compliant, pornographic, click-baiting, auto-redirecting or crash-inducing.

Google AdMob also has its Ad review center, but it only reviews ads sold by Google itself and does not cover ads from third-party networks. There are also independent third-party tools on the market such as AppHarbr and Boltive.

5. The AppLovin vs. Unity case is critical to AppLovin’s results this quarter. Questions we need to resolve going forward include:

If Unity did collect data, was this reflected in AppLovin’s 2Q26 miss and in Unity’s rapid IAA share gains in 2Q26?

AppLovin gave its 3Q26 guidance before it sent the cease-and-desist letter. Did its 3Q26 guidance implicitly assume that the Unity competition would be resolved?

Will a more competitive Unity model raise the AppLovin DSP’s bidding costs and in turn affect AppLovin’s take rate?

After Unity stops collecting data, will the gap between AppLovin AXON and Unity Vector widen again?

II. Publisher feedback on the impact on AppLovin’s acquisition costs

Around these questions, in this round of research we spoke further with five publishers, following up on changes in AppLovin’s and Unity’s bids, impression share and competitive dynamics since Q2, with a focus on the impact on AppLovin’s acquisition costs and take rate.

Before reading the research details, some technical background is needed. AppLovin’s take rate = (advertiser spend - cost paid to publishers) / advertiser spend. If Unity’s bids in MAX rise, it implies that improvements in Unity’s algorithm may make it more competitive when competing for inventory. If AppLovin’s MAX bids also accelerate y/y while Unity’s bids rise, it implies AppLovin has had to raise its bidding costs to protect share. This would also lower the take rate.

Therefore, DSP shares in MAX, and the y/y change in AppLovin’s MAX bids, indirectly reflect changes in AppLovin’s take rate.