Fighting Disinformation: How AI Detectors Combat Fake News Online

One of the most significant concerns surrounding the rise of artificial intelligence (AI) technology is that of disinformation. Even before the emergence of AI, the general public had begun to express the sentiment that much of the news that could be found online is either fake or dishonest, but this new technology has only compounded these worries.

What fake news existed before AI, at the very least, had to be deliberately constructed by a human being or group of people, but AI has automated the process of creating a misleading, or even harmful, narrative. As a result, many have resorted to using an AI Detector to identify AI-generated content.

The Potential Scale of AI-Generated News 

Large media organizations such as the New York Times reportedly publish about 200 pieces of journalism daily, but this pales compared to AI’s capabilities, which can generate hundreds or thousands of articles daily.

As if AI-generated writing was not enough, chatbots, image generators, and deep-faked voices can create news and news sources that appear legitimate and human-made. The average person may like to think they can identify AI-generated text, voice, and video. Still, a catchy headline and a striking image easily sway them.

Combating Fake News With an AI Detector 

Fortunately, the AI detector has emerged alongside the spread of AI-generated disinformation, providing users a resource to protect themselves from fake news online. While AI becomes increasingly difficult for humans to distinguish from legitimate news, and large media organizations struggle to identify genuine imagery and writing, AI detectors use the same technology to combat the spread of information.

How AI Detectors Identify AI-Generated Content 

An AI detector employs AI technology against AI-generated content, analyzing text, images, frames, or voices to determine whether AI created something. The most prevalent form of AI detector combats AI-generated text, utilizing machine learning (ML) and large language models (LLMs) to analyze the structural and linguistic features of a given piece of text.

Rather than mimicking human writing, an AI detector’s system is trained to compare and contrast human written work with AI-generated text. The two primary factors AI detectors use to identify AI-generated text are perplexity and burstiness, distinguishing human writing from AI writing.

Perplexity 

Perplexity simply refers to a text’s unpredictability, how likely it is to confuse the reader or hinder their understanding of a text. Humans write with greater word choice and sentence structure variance, while AI is streamlined and consistent.

As such, AI-generated text has “low perplexity,” while human writing has “high perplexity.” If a text is repetitive in word choice and sentence structure, it is more likely to be AI-generated.

Burstiness 

Burstiness is also a measure of variation in sentence structure but includes considering sentence length. An AI detector will analyze the length and structure of a text’s sentences to determine burstiness, with AI being more consistent in length and structure and humans tending toward variability. If a text consistently uses sentences of the same or similar length, it is more likely to be AI-generated.

AI detectors also utilize plagiarism detection software in their AI detection processes. All AI-generated content is based on existing text, and as such, it is more likely to plagiarize or make nonsensical references or do not exist in the first place. By integrating plagiarism detection, many AI detectors are equipped with another means for identifying AI-generated writing.

AI-Generated Images 

AI-generated images followed shortly after AI-generated writing, but the technology has already advanced rapidly. While it remains easier to identify than AI-generated writing, images can have a far more potent effect on one’s interpretation of online content. Fake news outlets may rely on AI-generated images to evoke a desired emotion in the reader, making the article seem more believable.

News imagery relies on credibility, so AI detectors must emerge to identify images and videos that are dishonest at best and harmful at worst. McAfee and Yahoo have developed a deepfake detector, an AI-powered solution for identifying and reporting AI-generated images in the news. Similarly to AI-generated text detectors, an AI image detector is trained on authentic and AI-generated images to identify patterns consistent with AI.

A Barrier Against Disinformation 

An AI detector can be a powerful tool for identifying misleading stories online for the average user and legitimate news organizations. While not every article will be unbiased or entirely truthful, AI-generated text and accompanying images demonstrate that a human being did not consider the piece’s content, creating dishonest narratives aimed solely at fulfilling a request. An AI detector is a barrier between disinformation and the user, protecting their online experience and keeping them properly informed.

This said an AI detector is only as effective as the technology it relies upon. The notion that AI is not entirely indistinguishable from human work reveals that an AI detector will not always catch onto an AI’s inconsistencies. While an AI detector remains valuable for maintaining online safety and avoiding fake news, users should remain aware that AI technology is still developing and improving.

Since an AI detector is not infallible, relying on AI detection services that offer detailed and comprehensive reporting on their AI content analysis is essential. This way, individuals and businesses can determine for themselves whether an article or other piece of news seems reliable based on the input of the AI detector. Transparency scores and recommendations help to highlight potential AI content, combating disinformation online.

The Value of an AI Detector 

AI detectors are valuable tools in today’s online spaces. By identifying the patterns and inconsistencies of AI-generated text compared to human writing, they protect users from content that lacks verifiable authorship and legitimacy, enabling them to avoid harmful misinformation. As fake news becomes increasingly abundant, these resources also help to prevent harmful misinformation.

———-

Have you read?
Happiest countries in the world.
World’s Highest (and Lowest) Life Expectancy.
Best Countries to Hide Money.
100 most popular wedding song ever.