Article

4 Tips for Spotting AI-Generated Pics

Images created by artificial intelligence (AI) software often circulate without disclaimers to say they're fake.

Published April 16, 2023

This page is part of an ongoing effort by the Snopes newsroom to teach the public the ins and outs of online fact-checking and, as a result, strengthen people's media literacy skills. Misinformation is everyone's problem. The more we can all get involved, the better job we can do combating it. Have a question about how we do what we do? Let us know.

AI-generated content is text or images created by artificial intelligence (AI) software based on text prompts input by a human user. Machine-learning algorithms enable the software to build databases of linguistic and visual models it can use to simulate human thought processes and creativity.

An AI system is likely to recognize the names, definitions, and likenesses of a wide array of common objects (say, canoes), historical figures (say, George Washington), and famous artists (say, Vincent Van Gogh), for example, and be able to create an image based on a prompt combining those words. This is a picture generated by the open-access AI image generator DALL-E based on the prompt "george washington in a canoe as painted by van gogh":

Similarly, an AI-text generator could be prompted to produce an encyclopedia entry about George Washington, or an AI chatbot (such as ChatGPT) could interact with a user to simulate a conversation about Washington's role in the American Revolution.

Though AI-content generators can be glitchy, and their output can be easily detectable as imperfect simulations, they have gotten better over time. Critics of artificial intelligence have raised alarms over the potential for plagiarism and the use of realistic "deepfake" imagery that could lead to widespread, and potentially harmful, misunderstandings.

For example, in the days surrounding Manhattan prosecutors' move to criminally charge former U.S. President Donald Trump, the Snopes newsroom saw a wave of AI-generated content on social media with Trump as its main character. 

With varying views on the case, AI-software users posted all sorts of fictional scenes featuring Trump or his rivals — from imaginary clashes between Trump and law enforcement officers to jail booking mugshots to depictions of his brief stay in New York last week for his arraignment. The images seemingly fooled some onlookers into thinking they were real.

Snopes does not want that to be you.

4 Tips For Identifying AI-Generated Images

If you're ever unsure whether a photo is a product of AI or if it's authentic, consider these things:

1. Zoom in. If the image appears in a social media post — for instance, if it's displayed underneath words in a tweet — click on the image specifically so it appears larger on your phone or computer. The more you can zoom in on an image's details, the better.

2. Hands and mouths (often) tell all. If the photo shows people, look at their hands or mouths. Do they have six fingers? Do they have two upper lips? These are details that general-use AI technology does not always get right (at least for now).

3. Reverse-image search. Save the image to your computer and perform a reverse-image search. If the search doesn't return any results (in other words, if there are no other places online where this photo appears), that's a sign that it could be a digital creation of some kind.

4. Snopes may have a fact check. Check snopes.com. We're continuously fact-checking these types of images, and we have a database of all kinds of suspicious photos you may find helpful. If you don't find what you're looking for, send the suspicious image our way. We'll try to look into it.


See also from Snopes:

David Emery is a West Coast-based writer and editor with 25 years of experience fact-checking rumors, hoaxes, and contemporary legends.

Jessica Lee is Snopes' Senior Assignments Editor with expertise in investigative storytelling, media literacy advocacy and digital audience engagement.