Skip to Main Content

Generative Artificial Intelligence (GenAI)

This guide provides an introduction to Generative Artificial Intelligence, originally created in Fall 2023 and updated in Winter 2024 by Librarian Greg Bem. Content based on Fall 2023 and Winter 2024 workshops developed by Greg Bem.

Evaluating Images for Authenicity

Questions to ask:

  • Who is posting this image? Is this a reliable source?​

  • What is the purpose of the image? Why is it being posted?​

  • Who is the intended audience?​

  • When is it being posted?​

  • Can it be verified or found elsewhere? ​

  • Can I find images from the moments before or after?​

AI image generation continues to improve. In addition to becoming familiar with common issues in AI generated images, there are different online tools that can help verify the authenticity of an image found online. 

These tools are not perfect. They shouldn't be relied upon completely, but they can be added to your "toolkit" of approaches to use when considering the authenticity of an image.

Reverse Image Search

A reverse image search can show where else on the internet an image can be found. This can help you evaluate whether the image is real by giving more information and sources to investigate for authenticity. 

  • Google Lens: On a computer, click the small camera icon in the Google Images search bar to search by image. On a smartphone, the Google Lens app will offer the same functionality. 
  • TinEye 

AI Detection Tools

Use AI to spot AI? There are a number of websites utilizing AI technology to help identify AI generated media. These can be helpful, but it is good to remain skeptical and not rely solely on this one tool. 

  • SightEngine: This tool evaluates based on the pixel content in an image and claims to be able to identify authenticity of images from popular Ai image generators. 
  • Hive Moderation: A Chrome extension that can process images, text, audio, and video. 

Metadata

Metadata is information about a file that is hidden (embedded) into the file itself. Image files typically contain EXIF data. This can include information on the type of camera or lens, GPS location, time stamps, and copyright details. This information, or the lack of such information, can be helpful in evaluating if a photo was taken by a camera or generated online. However, it should be noted that metadata can be removed from files, so again, this method should be used in addition to other approaches to evaluating image authenticity. 

AI images are relatively easy to generate and access. They may be useful for personal, professional, or even academic purposes (but must be properly cited!). However, there are other options for easily finding copyright-free or limited copyright images. 

Freely Licensed Image Sites

Library Image Databases

Real or AI? Image Evaluation Practice

Practice your AI image spotting skills with these interactive online quizzes.

AI Image Examples

An AI generated image showing a teacher in front of a class. The prompt used to generate the image is also included on the slide: "“create a photorealistic image of a group of adult students learning in a classroom. include a presentation in the background of the image.”, ChatGPT DALL-E 3, April 2025". Irregularities in the image suggest it is AI generated.

Classroom Instruction

  • Consider the hands on each person. Are there any issues in these details? In the bottom right corner, do you see anything that doesn't quite make sense? 
  • Would an instructor be likely to use the slide that is shown? 
  • What is on the laptop screen? Does it look realistic?

An AI generated image of a living room with large sliding glass doors looking out to the beach and ocean. The slide includes the prompt used to generate this image: "“create a detailed photorealistic image of a beautiful, large living room in an expensive beach house. the image should look out a large window to the beach and ocean beyond”, ChatGPT DALL-E 3, April 2025​"

Beach House Living Room

  • Consider the small details of this photo. Do the textures seem realistic? What about the bookcase? Do these look like real books?
  • What is happening with the fireplace? What about the door handles for the sliding doors?
  • Do the shadows all line up with each other and make sense for the lighting source?

An AI generated image of downtown Spokane, Washington, with the prompt used to generate the image: "“create a photorealistic image of a busy street in downtown Spokane, Washington”, ChatGPT DALL-E 3, April 2025​". The image has irregularities that inform the viewer it is AI generated. Image irregularities include: Incorrect/hallucinated signage. Inaccurate street lane markings. Inaccurate license plates and signage for PF Chang's.

Downtown Spokane, WA

  • Is this how downtown Spokane looks? Is there anything included in the photo that isn't really downtown?
  • How about the writing on the building on the right? Is that accurate? 
  • What about the license plates? And, is this a two way street or a one way street? Do the street lane markings make sense for the cars included?
  • How does your personal knowledge and experience with downtown Spokane influence your ability to identify this as an AI generated image?

An AI generated image of downtown Spokane, Washington, with the prompt used to generate the image: “generate a photorealistic image depicting an earthquake hitting Seattle, Washington”, ChatGPT DALL-E 3, April 2025. Image irregularities are based in the knowledge that this event did not occur in Seattle. Additional irregularities include incorrect signs, building damage not aligned with event, and issues with the traffic lights.

Earthquake in Seattle, WA

  • Did this event happen in Seattle?
  • Do the signs and traffic lights make sense?
  • Does the damage to the buildings align with the damage to the road? Would this damage be accurate for the type of disaster depicted?

An AI generated image of a living room with large sliding glass doors looking out to the beach and ocean. The slide includes the prompt used to generate this image: “generate a period-consistent photo of Abraham Lincoln”, ChatGPT DALL-E 3, April 2025. Image irregularities include issues with the buttons and button holes on the suit. The pocket watch chain does not make sense. The quality and cleanliness of the hair is inaccurate for the time period. The detail on the beard coloring is inaccurate for camera technology in this time period.

Abraham Lincoln

  • Take a look at details including the tie, buttons, and pocket watch. On close inspection, do these seem realistic? Do they make sense?
  • How about the hair? Does the cleanliness and texture of the hair make sense for this time period?
  • Would camera technology at this time be able to capture details like the graying of the beard?

Image of a slide with text: Common AI Image "Tells" Text and numbers Inconsistencies in details (hands, fingers, nails, teeth, ears) Lack of detail Skin and hair texture Issues in patterns Unnatural proportions or shadows Strange accessories or context mismatches Inconsistent or improper reflections "Vibes" (something just feels off)

Common AI Image "Tells"

Review this list of some common visual indicators an image may be AI-generated. Note: This is not a complete list, and models are constantly improving.