Conceptual image of artificial intelligence creating an image.
As the volume of AI-generated content continues to proliferate and grow online, so does our collective need to successfully investigate “fakes from facts.” This article is the first in a three-part series that will provide some easy tips, tactics, and tools that anyone can use to sharpen their eye for AI-generated content.
Today’s first installment looks at five ways to tell if an image has been created with artificial intelligence.
Texture and Resolution: Images generated with artificial intelligence may show inconsistencies between textures. Look for areas of unnatural smoothness or blur where there should be continuous texture patterns.
This visual nit usually occurs due to the processing limitations of the AI, as it has difficulty rendering images uniformly with current technology and its programming. These hues are usually evident in the background, middle, and side frames of the image — not so much in the main focus areas.
Anatomy errors: While this particular area has seen rapid improvement, the AI still makes mistakes when it comes to visual proportions of hands and feet, fine details around toes, fingers and teeth as well as precise feature placement of the face.
These errors can range from subtle to glaringly obvious, but are indicative that the image has been edited or created using AI technology. If there’s something about the people in the image that doesn’t look right to your eyes, but you don’t know exactly what it is – it’s probably done with AI.
Tool suggestions: As the wave of deepfakes rises on the webs, there are several AI image detection tools available. Two that you can trust without any reservations are Scope and Cell. Both have free versions of their AI image detection feature, as well as multiple advanced features beyond just imaging, for a fee.
What’s really cool is that they both give you a percentage indication as to which specific AI platform was used to produce the image, such as Midjourney, Dall-E, Firefly…etc. In the case of the image below, Sightengine pinpointed with 99% confidence that the purple dinosaur was holding a bitcoin while saying “HODL enabled” was actually created using the Midjourney genAI image platform.
In a recent comparative study Conducted by the University of Rochester and the University of Kentucky, Sightengine was determined to have the highest accuracy on the current market when it comes to AI detection.
The Sightengine interface shows 99% confidence that the author of the article used Midjourney AI … [+]
However, it’s worth noting that no tool is foolproof, so using multiple tools is best practice. The AI-generated dino image below was also tested on Hive, which accurately detected that the image was generated using artificial intelligence with a confidence score of over 99%.
Hive’s AI detection interface accurately shows that the image the article author uploaded was … [+]
Shadows and Lighting: This is another area where the current generation of genetic AI seems to struggle. It struggles to accurately recreate the contours produced by real lighting and shadows. Even to a casual observer, these visual anomalies won’t line up with contrast images for a lifetime and can be a clear tip that the image is fake AI.
Specific things to look for might be shadows that are too dark or too light relative to the ambient light in the image, or shadows cast in directions that don’t match the directional light sources within the image.
AI Watermarking: Several of the artificial intelligence tools available in the market automatically place a subtle watermark or tint pattern on the free versions of the images produced.
This serves as an incentive to increase sales in the paid version of the service, which produces images without the watermark. Regardless, AI watermarks are one of the easiest things to spy when trying to identify whether an image is real or not.
As AI technology continues to evolve – rapidly – the tools for detection must keep pace because soon the human eye and intuition will not be enough. As you evaluate which tools you want to consider using in the future, here are some things to keep in mind when it comes to AI detection:
- Understand potential inaccuracies and limitations in current detection methods.
- Always consider the source and context of the image in question.
- Make sure you use multiple tools for cross-validation purposes.
- Stay up to date with the latest developments and updates in the AI crawl space.
There is no doubt that as the fields of creating and detecting artificial intelligence continue to advance, new and improved tools will appear on both sides of this ledger. We can only hope they remain in balance.
The next installment will explore tips, tactics and tools for spotting AI written content.
