Fake websites and AI technologies: what you need to know to avoid falling for scams

Virginia Tech experts examined three different aspects of the spread of AI-generated fake news websites and efforts to combat them.

How technology helps create and detect fake news

Professor of engineering and machine learning Walid Saad explains that the ability to create websites with false information has existed since the early days of the internet and predates the AI era. However, with the advent of AI, it has become easier to analyze large amounts of data and generate plausible texts. LLMs allow attackers to generate content that appears to be accurate information, making such websites particularly dangerous.

He emphasizes that these sites will continue to function as long as people continue to consume and share content from them. The spread of disinformation on social media motivates creators of fake websites to keep operating.

Solving this problem requires joint efforts by people and technology. AI not only contributes to the spread of fakes but can also help in their detection. However, human input—both from readers and administrators—is extremely important. Users and news agencies are responsible for not amplifying the spread of false information, and user reports of erroneous information help improve AI-based detection tools.

At the same time, measures to combat disinformation must respect the First Amendment to the U.S. Constitution and not hinder freedom of speech.

What laws can and cannot achieve

Casey Myers, an expert in communication policy, notes that regulating political disinformation raises many practical and legal issues. Although many countries have recognized the need for action ahead of the 2024 elections, the accessibility of AI makes creating disinformation, including deepfakes, easier than ever.

Legislation faces difficulties—many creators of fake content remain unidentified, and if they are in other countries, it is even harder to hold them accountable. Additionally, technologies like Sora demonstrate that high-quality AI-generated content is becoming increasingly accessible for mass production, and even watermarks or labels can be removed or altered.

In the U.S., social media websites hosting political disinformation, including deepfakes, are protected from legal liability under Section 230 of the Communications Decency Act. Therefore, the fight against disinformation often comes down to the platforms' own rules, which has drawn criticism and raised questions about bias. There is also discussion about holding AI-creating platforms accountable, but there is currently no reliable system capable of completely preventing the creation of AI-generated disinformation.

How to protect yourself from disinformation

Julia Ferrar, a librarian and digital literacy instructor, provides advice on how users can recognize false materials themselves.

She advises looking at where the publication comes from: from a professional, authoritative source or from an unknown website. If there are doubts, it is worth opening a new tab and quickly finding information about the site or checking if there is an article about it on Wikipedia. This is called lateral reading—seeking information outside the text itself and verifying whether other reliable media report the same news.

Other tips:

— fake materials often appeal to emotions—pause and think if something provokes a strong reaction

— check headlines and images using fact-check searches

— a very generic website name can be a sign of AI-generated news

— sometimes you can see error messages left by AI tools in the text if the text creation violated usage policies—this can be a red flag

— when analyzing AI-generated images, experts point out unusually perfect pictures, as well as strange hands or feet.