AI-Generated Short Videos and Malicious URLs on the Rise: Why Are Sellers Becoming New Targets of Online Scams, and How Are Algorithms Reshaping Fraud and Cybersecurity?

The rapid development of artificial intelligence (AI) is fundamentally transforming the landscape of online fraud. Generative AI tools have significantly reduced the time and cost required to create convincing scam content, enabling fraudsters to produce realistic videos, short-form clips, and virtual personas. Combined with algorithm-driven content distribution and analysis of users’ digital footprints, these technologies allow fraudulent messages to reach specific audiences with unprecedented precision. Online scams are no longer limited to ordinary consumers. Sellers, merchants, and businesses whose contact details and activities are publicly accessible are increasingly becoming targets. As digital fraud grows more sophisticated, maintaining digital trust requires an integrated approach that combines technological innovation, cybersecurity awareness, domain name management, and cross-sector collaboration to establish more comprehensive safeguards.

Chia-En Lee, Head of the Domain Name Division at the Taiwan Network Information Center (TWNIC), and Che-Wei Liang, Founder and General Manager of Jishiduo, share their frontline observations of emerging scam tactics and practical experience in developing effective anti-fraud mechanisms. Their perspectives highlight how AI is changing the way scams operate, why individuals and businesses remain vulnerable, and how government agencies, research institutions, and private companies can work together to strengthen the broader digital security ecosystem.

One of the most significant changes in online fraud is the shift from mass messaging to highly personalized deception. In the past, fraudsters typically relied on sending large volumes of similar messages. Today, generative AI and data analytics algorithms can tailor fraudulent content to individuals’ interests, concerns, and everyday circumstances. Lee points out that although scams continue to exploit fundamental human desires and fears, fraudsters are becoming increasingly adept at using publicly available online information, including search histories, browsing behavior, and social media activity, to make their approaches more persuasive. Social engineering attacks may also exploit victims’ personal networks. Once someone clicks a malicious link, their account or identity may be used to contact friends, colleagues, and other connections, allowing the attack to spread even further.

Liang notes that AI-powered video production tools have lowered the barriers to creating virtual personas and maintaining false identities. Scam content has evolved beyond static images and text to include short-form videos, social media stories, and virtual presenters. AI-generated characters can be shown engaging in everyday activities, creating an appearance of authenticity that helps establish viewers’ trust. As these videos accumulate likes, comments, and other forms of engagement, they may appear even more credible. Consequently, it has become increasingly difficult for ordinary users to determine whether content is genuine based on appearance alone.

At the same time, fraudsters are increasingly targeting online sellers and businesses whose contact information and commercial activities are publicly available. Lee explains that fraudulent websites often imitate legitimate domains by substituting visually similar characters. For example, two lowercase “v” characters may be used to resemble a “w,” the letters “rn” may imitate an “m,” or the number “0” may replace the letter “O.” These subtle differences can deceive users who do not carefully inspect website addresses. Fraudsters may also send counterfeit government notices, domain expiration reminders, or email verification requests to create a sense of urgency, prompting recipients to click malicious links and enter their account credentials or credit card information.

Liang further observes that fraudsters are increasingly taking the initiative by posing as prospective buyers and contacting sellers directly. Exploiting sellers’ desire to complete transactions successfully, scammers may insist on using specific payment or logistics services and direct sellers to counterfeit websites. They may then claim that identity verification has failed or that a problem has occurred with the seller’s bank account, persuading the seller to follow additional instructions. Ultimately, the victim may unknowingly disclose login credentials or transfer funds to the fraudster. These cases demonstrate how seemingly routine business interactions can be turned into opportunities for financial fraud.

The growing sophistication of deepfake technology presents another major challenge. Neither visual inspection alone nor reliance on a single detection tool can guarantee the accurate identification of AI-generated videos. Liang emphasizes the importance of examining multiple indicators, including body movements, facial details, and consistency in lighting and shadows. Earlier generative tools sometimes produced obvious anomalies, such as unnatural hand movements or incorrect numbers of fingers. More advanced face-swapping technologies, however, can eliminate many of these visible flaws. Detecting suspicious content therefore requires more sophisticated analysis, including assessing whether facial shadows and changes in lighting are consistent with the subject’s movements and the physical properties of the surrounding environment. Combining multiple AI detection algorithms can improve reliability, but no single method currently delivers perfectly accurate results.

Beyond individual vigilance, domain security and proactive monitoring are also essential to preventing online fraud. TWNIC uses monitoring and early-warning mechanisms, promotes its Green Domain Certification, and applies Response Policy Zones in the Domain Name System (DNS RPZ) to help identify and block malicious websites. Lee explains that when major public policies or initiatives attract widespread attention, such as government cash-distribution programs, TWNIC works with government agencies to monitor suspicious domains and proactively identify URLs that closely resemble legitimate websites. These efforts help reduce the risk of fraudsters exploiting public interest in government initiatives to distribute false information through counterfeit websites.

Green Domain Certification offers another way to strengthen digital trust. Businesses that complete identity verification can display the certification mark in designated areas of their official websites, helping users identify verified websites and reducing the risk of visiting fraudulent imitations. Encouraging the use of Taiwan’s .tw domains can also help businesses benefit from locally available identity verification and security services. Together, these measures, supported by cooperation among government agencies, international cybersecurity providers, and domain management organizations, can improve the efficiency of malicious website monitoring and strengthen coordinated responses across institutions.

Liang emphasizes that effective anti-fraud efforts also depend on timely reporting and the appropriate use of data. Individuals and businesses who encounter suspicious messages or fraudulent URLs can report them through the relevant channels provided by Taiwan’s Ministry of Digital Affairs. These reports not only help authorities investigate and remove malicious content but also provide valuable data for training AI-powered anti-fraud models. Liang also shares that Jishiduo participates in the Ministry’s anti-fraud initiative known as the “Good People Team” , using reported cases to train AI models that automatically detect scam posts and accounts on social media. Through collaboration with the Industrial Technology Research Institute (ITRI) and government agencies, these initiatives aim to improve detection efficiency while reducing the manpower and resources required to process fraud reports.

Ultimately, the rapid evolution of AI-enabled scams demonstrates that no individual, organization, or single technology can address the problem alone. Effective protection requires an integrated strategy that brings together cybersecurity awareness, reliable domain management, advanced detection technologies, timely reporting, and coordinated action across sectors. By combining the expertise of government agencies, research institutions, cybersecurity organizations, and innovative private enterprises, Taiwan can strengthen its digital security infrastructure and improve its ability to respond to emerging threats. Deepening public-private collaboration will remain essential to building a more resilient anti-fraud ecosystem and preserving trust in an increasingly interconnected digital environment.