AI-Powered Ad Fraud Surge: DoubleVerify Flags Massive Spike in Malicious Apps

A concerning trend identified by the DV Fraud Lab reveals a dramatic rise in fraudulent mobile apps available for download, both on iOS and Android. These are lower quality apps that largely exist to serve ads or engage in other malicious activity. In many cases, they are powered by or generated using AI, with bad actors creating shell operations to hijack Android/iPhone/iPad devices and/or falsify millions of fake impressions. 

No single provider is alone in feeling the impact of AI-powered fraud. In fact, major app stores across mobile and Connected TV (CTV) are seeing similar trends. In 2024, the DV Fraud Lab uncovered the impact of GenAI-infused fake reviews on the ecosystem, as covered by Fortune. Previously, in 2023, Roku and DV partnered to shut down “SmokeScreen” – a malware scheme that targeted CTV devices. Now, the growing number of fraudulent mobile apps shows that all marketplaces must remain vigilant against this acceleration. 

DV detects ad fraud at scale, across all environments, allowing these fraudulent apps to be immediately classified for ad impression fraud. So far in 2025, DV has already classified nearly three times more fraudulent iOS apps in comparison with the average volume over the last five years. An even more significant increase was measured for Android apps, with nearly six times more fraudulent Android mobile apps in comparison with the average of the last five years. This explosive growth in fraudulent apps highlights the need for advertisers to work closely with third-party ad verification and media effectiveness providers to avoid wasted ad spend. In the GenAI era, quality standards for mobile apps are becoming increasingly difficult to maintain.

For context, just one year ago, Apple reported rejecting more than 1.76 million app submissions for failing to meet the App Store’s stringent standards. This year, that number has risen to 1.93 million — almost a 10% increase. Could AI be to blame? 

AI: The New Tool for Scammers

One of the key drivers behind this surge appears to be the use of AI. Fraud schemes are leveraging AI not just to generate fake traffic, but also to build more convincing and realistic-looking apps – making it harder for marketplace reviewers to identify and reject them at scale. These apps can also simulate legitimate user behavior, making the apps difficult to identify using traditional fraud-detection methods.

In the example below, a gaming app includes a detailed text description, promising users a thrilling experience once they install the game. In reality, many of the reviews found for the app appear inauthentic. For example, DV’s analysis found multiple app reviews containing repetitive language that does not match the nature of the app. This includes 5 star reviews referring to the app as “professional software” or “Contains many problems, I like.”

Another example is an app that was used by a scheme that produced over 200,000 daily impressions, causing significant monetary damage to unprotected advertisers. The app had minimal graphics and nearly zero reviews, making the significant number of impressions appear suspect. However, the app managed to evade App Store detection and remain active, in part by creating a facade of reliability using LLM produced language and a template site.

The template site of the app’s developer was hosted on tempisite.com, a service that has been repeatedly used by several scams and phishing schemes. For example, the following screenshot shows that a phishing site impersonating Facebook is hosted on tempisite.com:

Source: https://gridinsoft.com/online-virus-scanner/url/facebook_support-tempisite-com  

This connection between different types of attacks – such as phishing scams and adtech schemes – is another characteristic of fraud in the GenAI era. Prior to the emergence of GenAI, sophisticated fraud used to be the exclusive domain of fraudsters who had the time and expertise to access darknet forums and find resources to host their operations. This is no longer the case. With GenAI, a resource like tempisite.com can become well known even to amateur fraudsters, allowing them to successfully create a wide variety of malicious attacks much faster than ever before. Thus, with GenAI, the opportunity to commit fraud across many industries and sectors has exploded.

Furthermore, fraudsters now employ device emulators and manipulation tools to mimic iOS device traffic, techniques that were once predominantly associated with Android environments. This shift marks a dangerous new chapter in mobile ad fraud, where even Apple’s historically stringent ecosystem could be penetrated with increasing ease.

Gatekeeping in Question for Mobile and CTV App Stores

2025 is proving to be a critical stress test for app marketplaces. DV has also observed a sharp rise in classifications of other types of low-quality or abnormal traffic from these apps, suggesting a broader systemic issue.

“There’s a noticeable increase in apps slipping through major marketplace review processes that were previously unlikely to be approved,” says Yuval Rubin, Fraud Detection Group Lead at DV. “This could point to a combination of more sophisticated fraud tactics.” 

Buying Direct Is Not the Magic Cure

In the wake of growing fraud concerns, many marketers are leaning heavily on direct buys and Private Marketplace (PMP) deals, assuming these channels offer a cleaner, more transparent source of traffic. However, experts warn that this assumption is increasingly flawed. AI-powered fraud is not limited to open exchanges; it’s infiltrating every buying method, including direct and PMP deals. 

DV data indicates that many traffic sources marketed as “direct” have been discovered to be reselling invalid or misrepresented impressions, often stemming from the very kind of deceptive apps discussed earlier. In its recent releases — including AI Slop Websites, Mobile Fraud on Sleep Apps and CycloneBot and FMScam — DV has emphasized that fraud schemes proliferate across all parts of the industry, impacting web, CTV, mobile, audio, gaming, retail media, and more. With AI fueling this widespread fraudulent activity, the message is clear: No buying path is immune, no channel is a safe haven, and trust in the supply chain must continuously be audited and verified rather than assumed.

What Happens Next?

Should this trend continue, it would have broad implications for advertisers, developers and end-users. Advertisers will need independent measurement to validate that their dollars are well spent and effective when buying in-app inventory, and developers of legitimate apps may suffer from increased scrutiny or stricter policies. For all marketplaces, it’s a critical moment to assess review mechanisms, transparency initiatives, and investment in AI-powered fraud detection.

As always, DV continues to vigilantly protect the industry. Each year, the DV Fraud Lab employs a combination of AI (including GenAI) and expert analysis to identify thousands of fraudulent apps and sites, as well as dozens of web, in-app, and streaming fraud schemes. Thanks to sophisticated tools and algorithms used to accurately identify individual invalid impressions, DV stays at the forefront of emerging fraud techniques and provides maximum protection to customers. 

Should you have questions about this fraud scheme or DV fraud safeguards in general, please reach out to your DV account manager or contact us on our website.