As AI-generated content proliferates online, discover the leading tools and technologies designed to … [+]
Adobe Stock
Since the launch of ChatGPT just two years ago, the volume of synthetic – or fake – content online has increased exponentially.
Firstly, not all “fake” content is inherently bad. Generative AI text, image and audio tools have streamlined many repetitive tasks, from drafting routine letters and notices to storyboarding and prototyping in more creative projects.
But AI-generated content becomes problematic when it’s intended to mislead, misinform or spread fake news. Some have even gone as far as to say it threatens to destabilize democratic processes and create a “truth crisis.”
So what can be done? Well, luckily, a number of methods of differentiating between AI-generated and authentic content have been developed. These include sociological approaches, emphasizing the importance of education and critical thinking. They also include technological solutions, often leveraging the same generative and machine learning models used to create “fake” content, repurposed to detect it
instead.
Here, I’ll focus on the latter. I’ll start by covering how they work, then take a look at some of the most popular tools and applications in this category.
Deep Fakes, Fake News And Disinformation
First of all, It’s important to distinguish between “fake news” and “fake content”.
The last few years have seen a big increase in both. However, the term “fake news” covers any deliberately constructed stories, lies, or disinformation designed to deceive. Whereas “deepfake,” “fake content,” or “synthetic content” specifically refers to content that’s not just designed to deceive but also generated by AI.
This deception could simply be for the sake of entertainment – as in the case of viral internet fakes like “deepfake Tom Cruise”, or “Pope In A Puffer Jacket.”
On the other hand, and increasingly, it could also be intended to cause real harm, such as influencing elections, damaging trust in public figures, or spreading geopolitical propaganda.
This year, ahead of upcoming elections in many countries, the WEF recognized AI misinformation as the biggest cybersecurity risk facing society. This all suggests that developing methods and tactics for identifying and combatting the rise of deepfake content is important for all of us.
What Are AI Content Detectors And How Do They Work?
In the simplest terms, most AI content detectors work by analyzing content and attempting to spot patterns that suggest it may have been generated by AI.
Often, they rely on AI itself to do this, leveraging neural networks that are trained to recognize typical traits.
For text, this could be particular phrases or ways of structuring information that is typical of large language models (LLMs), such as those powering ChatGPT or Google Gemini.
With images, this could mean looking out for telltale mistakes. For example, it’s frequently observed that AI image generators will often have difficulty with adding the correct number of fingers to their drawings of hands, correctly rendering text, and dealing with lighting and shadows.
It’s important to remember, however, that even the best tools are not foolproof. For example, it’s easy to mix AI and human-generated content to create hybrid content, but that’s likely to confuse AI content detectors.
Because of this, most of the tools covered here don’t categorically determine whether content is either AI or genuine. Instead, they are more likely to assign a probability or estimate how much of the text is likely to be AI-generated.
(To demonstrate this, I fed the text of this article, which is entirely human-written, into all of the text-based AI detectors mentioned here. You can see the results below)
The Best AI Content Detectors
AI Or Not
This paid-for site detects the use of generative AI in both images and audio.
Copyleaks
AI text analysis that’s widely used by businesses and academia.
Is this article written by AI? No
Deepfake Detector
Identifies fake video and audio with a claimed 92% accuracy.
Deepware
Deploy professional-quality deepfake detection resources on-premises for businesses.
GPTZero
One of the first widely available AI text detectors.
Is this article written by AI? 4%
Grammarly
The real-time grammar-checking plugin also offers an AI content detector.
Is this article written by AI? 50%
Hive Moderation
Designed to provide real-time moderation of video, audio and text content, also detects AI content.
Is this article written by AI? 0%
Is It AI?
Machine learning-powered AI image detector with free and paid-for options.
Originality
This lets you verify that the content you are planning to publish is authentic and trustworthy by checking it for AI, as well as plagiarism and factualness.
Is this article written by AI? 3%
Plagiarismcheck
Powerful AI text detection suite, with specialized tools for educational use cases.
Is this article written by AI? 0%.
Quillbot
Free-to-use AI text checker with no sign-up requirements.
Is this article written by AI? 0%
Winston
Winston is a comprehensive AI checking tool that can detect fake images as well as text. It also offers a certification program, certifying content as human-created.
Is this article written by AI? 0%
As AI-generated content becomes increasingly sophisticated, the tools and technologies we use to detect it must evolve in parallel. While today’s AI content detectors offer valuable insights, they’re not infallible – as demonstrated by the varying results when testing this human-written article. The key lies in using these tools as part of a broader approach to content verification, combining technological solutions with critical thinking and digital literacy. As we navigate an increasingly complex information landscape, these detection tools will become essential components in our collective effort to maintain digital truth and combat harmful misinformation.
