The AI detector tools that can help you check content for plagiarism, fakes and scams
The rise of artificial intelligence (AI) tools like ChatGPT, Google Bard, and Jasper has caused a deluge of concerns from the global creative community and schools worldwide, which fear that the new technology could be used to replace them, steal their work, or cheat the system.
While many have called for prompt regulation of the sector before it’s used to potentially disrupt and even destroy our society, others have praised new tools for the ways they can help users in their daily lives - from deciding what to cook for a romantic dinner to helping them prepare for a high-stake job interview.
What’s clear is that AI technology has snuck into every aspect of our lives. Whether it’s an image shared on social media or the message you’ve just received on LinkedIn from a hopeful recruiter, it’s now normal to double-guess the origin of the content we see online.
But how do we know what content has been genuinely created by a human being, and what’s been manufactured by AI?
Here's a look at several online tools that can help you spot AI-generated content - whether it’s text, images or videos.
GPTZero
It is becoming increasingly important for school teachers and university professors to be able to tell the difference between an AI-generated text and a human-made one.
AI detector GPTZero was created to help these professionals identify content generated by ChatGPT, GPT4, Bard, LLaMa, and other AI models.
It works by analysing the text uploaded and scanning it for plagiarism against thousands of documents online, highlighting the parts of the document written by an AI model and those written by a human in different colours.
With the free "basic" plan, users can upload up to 10,000 words per month and do seven scans an hour, while there are more options available for those willing to pay for a monthly subscription.
Crucially, GPTZero specifies that it does not wish to be used to "punish students" but rather to give them a chance to receive assignments that cannot be solved with AI.
Copyleaks AI Content Detector
The company that created Copyleaks AI Content Detector has years of experience checking plagiarism, giving this tool a certain advantage over its competitors.
According to Copyleaks, the detector can spot AI-generated text with a 99.1 per cent accuracy - which makes it without a doubt the best tool currently available.
It’s also the only one which is constantly updated for new AI models and supports multiple languages other than English.
The tool allows you to paste any text above 350 characters and have it analysed by the AI detector, which will then come to a definitive conclusion on whether the document contains human text or whatever percentage of AI-generated content.
SynthID
Google has recently developed SynthID, a tool that allows users to tell if an image or artwork is AI-generated by embedding in them digital watermarks which are imperceptible to the human eye, but detectable for identification by the app.
The tool is currently available for Google Cloud customers only, but it should eventually be expanded to more users.
As this tool isn’t an option for the wider public, here are a few tips for spotting an AI-generated image that does not involve using an online detector.
First, watch out for extra fingers, as AI is famously bad at drawing human hands, often adding a sixth finger or cutting off the thumb. Then, pay attention to the details: if anything is weirdly blurry when it doesn’t make sense that it should be, it might mean that the image is AI-generated.
Overly rendered, super-contrasted images that feel a little "uncanny" are often AI-generated, rather than badly post-produced. AI is also pretty bad at rendering text in images, which means that often any text in artificial images might be garbled or make no sense.
Deepware Scanner
Deepware Scanner can scan and detect deep fakes, as its name suggests, in videos, which include spotting manipulations and AI-generated content.
The website accepts "scan requests" for suspicious videos which will then be uploaded and scanned by Deepware for authenticity.
Other similar deep fake detection tools are Sentinel, Reality Defender, WeVerify, and Microsoft’s Video Authenticator Tool.
