Skip to Main Content

Research Guides

Artificial Intelligence for Image Research

A guide on how to use Generative AI for image generation, editing, concept creation and development.

Accuracy and Reliability


When using AI tools, it is important to be wary of the accuracy and reliability of generated content. Generally in AI images, wonky hands, and other obvious visual tells (smears, blurring, and unintelligible text), more dangerous glitch artifacts are only visible if you have enough background knowledge on the subject matter.

As an example, take this AI-generated image of a lady in a kimono vetted by a kimono expert:


At first sight, the obvious telltale sign that this image is AI generated is the hands.

Fingers are disproportionate and odd, slightly blurry and unclear.



Secondly, the draping of the bottom half makes no sense, draping over the wrong direction.


Third, there is this extra roll of fabric coming from the hem draping on the ground that shouldn't exist.


Finally, the undershirt of the collar is folded right over left, which is used only for the dead. 

Together, these oddities point to the fact that this image is AI generated, but is only discernable to a keen eye, and if the viewer possesses background knowledge on the subject. It is not only important to exercise a healthy amount of skepticism when it comes to images sourced from the internet. Malicious actors are already using AI as a means of generating inflammatory or politically-motivated imagery in order to create misinformation, and caution must be exercised when encountering these images.

Here's a fun test to see if you can spot the AI-generated image, in order to drive home how difficult a process this can be.

Furthermore, you can use tools such as this to test images and see if they are AI-generated.