AI generated images are becoming more prevalent on social media

HELENA — You may have been scrolling through apps like Facebook and come across images that just look too good to be true.

If users do not have a keen eye for detecting images created or enhanced by AI, they can be fooled into engaging with fake content, spreading misinformation in the process.

Some accounts post dozens of AI-generated pictures a day to farm engagement from users. Very limited measures are currently in place to detect these images on platforms like Facebook, Instagram, and X.

Jason Neiffer, the executive director of Montana Digital Academy, is currently working on implementing AI into online classes across the state. AI can be a tool when it comes to education, but some can abuse this tool for clicks on social media.

"The game on social media is to get your engagement and to get your click. It wants your like, share, or comment which is why a lot of the images on Facebook tend to be more of the dramatic and more of the fantastic, said Neiffer.

"We call that glance media. If something you look at and spend a second or two, maybe three if it catches your eye before we react or move on to another image and that's something really a product of the social media era," Neiffer continued. "We tend to scroll past things pretty darn quickly."

Glance media leads to decreased media literacy, or the ability to critically analyze a piece of media to determine accuracy and credibility.

I uncovered a study conducted by Facebook IQ back in February 2017 about creativity in posts on social media. In this study, they found that people spend an average of 1.7 seconds with any given piece of content on mobile devices.

After seeing this, I conducted my own study with our very own MTN Journalists to see if they could detect the difference between real photographs and AI-generated images. The catch was that they had only 1.7 seconds to do so, simulating an average scroll on social media.

Out of the 13 participants who took my 20-question quiz, only 6 got 80% or better, with an average group score of 71.9%.

Scores on the Quiz:

I also wanted to test how confident people were in their answers, given they only had a short window of time to decide. only 70 of the 260 questions asked were "very confidently" picked out as AI or not.

Confidence Level for Each Participant:

Even MTN Chief Meteorologist Curtis Grevenitz admits that AI is becoming hard to detect.

"AI has a bunch of amazing capabilities, but the fact that it can detract from what is real and what is really happening out there, that's dishonest to me," said Grevenitz.

He also added, "Artificial intelligence has gotten us all, it's gotten me. As far as weather goes, we don't need AI to generate images and other media because I think there's enough natural beauty that is occurring here in this world without it."

When scrolling on social media, make sure to stop for an extra few seconds to take a closer look at some images that seem too good to be true, or, if it's weather-related, consult your local meteorologist.

Here are a few tips to help detect AI-generated images when scrolling on social media.

• If the picture looks "too perfect", it probably is.
• AI generation often contains minor, noticeable attention-to-detail errors. Like blurry or made-up characters on the license plate of a vehicle.
• AI-generated images can have inconsistent lighting and shadows.
• Backgrounds can be overly simplistic or overly complex, making the depth of the picture look unrealistic.
• AI struggles with creating realistic, living creatures, including humans. They will oftentimes look cartoonish and miss a lot of complex body details.
• You can view an image's metadata, or information about where it was taken, when it was uploaded, what camera settings were used, etc. If you're on your computer, right-click the image and select “Properties” to view metadata. If you’re on your phone, access image details through the information icon, or through apps like Google Photos.