Pushpaganda Scam Hijacks Google Discover with AI-Generated News

Executive Summary

A sophisticated ad fraud campaign dubbed "Pushpaganda" has been exploiting Google's Discover feed and AI-generated content to propagate deceptive news stories, ultimately coercing users into enabling malicious browser notifications. According to research from Sekoia, the operation has generated over 10,000 articles across a network of more than 50 spoofed news websites, targeting millions of users with scareware and financial scams. The campaign represents a significant evolution in search engine poisoning (SEO) tactics, leveraging the trust and reach of a major Google platform to initiate fraud.

Technical Analysis

The Pushpaganda operation follows a multi-stage technical process designed to maximize reach and user deception. Threat actors first establish a network of fraudulent news websites, such as newscase[.]online and newsexplorer[.]online, which are designed to mimic legitimate news outlets. Content for these sites is generated at scale using AI, focusing on sensationalist topics like celebrity scandals, technology leaks, and natural disasters to attract clicks.

These AI-generated articles are then optimized for search engines and, critically, for inclusion in Google's Discover feed—a personalized content stream presented to users on the Google homepage and mobile app. Sekoia's analysis indicates the campaign successfully placed its malicious sites into Discover, granting them massive, unsolicited visibility. When users click these links, they are presented with a fabricated "breaking news" video player and a prompt urging them to "Click Allow to watch the video." Granting this permission subscribes the user to persistent browser notifications from the malicious domain.

Once subscribed, the notification channel is weaponized. Users are bombarded with alerts that direct them to scareware pages falsely claiming their device is infected, or to fraudulent investment schemes and phishing sites. The infrastructure uses a chain of redirects through ad networks and traffic distribution systems (TDS) to obfuscate the final scam destination and profile the victim.

Tactics, Techniques & Procedures

The campaign employs a clear sequence of techniques:

1. Resource Development (T1583.008): Establishment of a network of over 50 fraudulent news domains using generic names (e.g., newscase[.]online).
2. Generate AI-Generated Content (T1659): Use of AI to mass-produce convincing, sensationalist news articles to attract organic and platform-driven traffic.
3. Search Engine Optimization (SEO) Poisoning (T1608.005): Optimization of AI content to rank in search results and, pivotally, to qualify for inclusion in Google's Discover feed.
4. Deception (T1657): Use of fake video players and urgent prompts to socially engineer users into enabling browser notifications.
5. Abuse of Notifications (T1655): Leveraging the subscribed notification channel to deliver malicious links directly to the user's desktop or mobile device, bypassing traditional web navigation.
6. Traffic Distribution System (TDS) Use: Implementation of redirect chains to filter traffic and direct users to tailored scam endpoints based on geolocation or device type.

Threat Actor Context

The threat actor behind this operation is tracked by Sekoia under the name "Pushpaganda." While specific attribution to a known group is not provided, the campaign's primary objective is financial gain through ad fraud and scam affiliate programs. The operational sophistication—combining AI content generation, platform abuse, and notification hijacking—suggests a technically capable group focused on large-scale, automated fraud rather than targeted intrusion. The use of AI for content creation at this scale is a notable evolution in the threat landscape for spam and SEO poisoning campaigns.

Mitigations & Recommendations

Organizations and individual users can take several steps to mitigate this threat. For system administrators and security teams, blocking the known malicious domains at network or DNS filtering layers is advised, though the campaign's use of a large, evolving domain list limits the effectiveness of static blocks. User awareness training is critical, emphasizing that legitimate sites rarely require enabling notifications to view basic content. Users should be instructed to deny notification prompts from unfamiliar websites and to regularly review and remove unwanted notification permissions in their browser settings (typically under Settings > Privacy and Security > Site Settings > Notifications).

For platform providers like Google, the incident underscores the challenge of automatically vetting content for deceptive practices within algorithmic feeds like Discover. While specific platform-side mitigations are not detailed in the source material, the research highlights an ongoing cat-and-mouse game between fraudsters and feed algorithms.