Thing 6

Thing 6. Detecting Deepfakes

Purpose – Spot the Deepfake

Now that we have everyone on alert for deepfakes, what can we do to detect those that aim to spread misinformation or are malevolent in their intent? Or even those that are entertaining.


Most of us don’t have a problem identifying cheapfakes. Spotting the bunny ears and nose on that Snapchat shot isn’t particularly difficult. Spotting the machine-learning created deepfakes videos or audio can be more of a challenge. While the threat of deepfakes is yet to come, according to experts, it is worthwhile to learn to spot deepfakes as a guard against the spread of misinformation and its malignant intent.

Why be concerned about deepfakes? Here are a few ways deepfakes can be used for malignant purposes:

  • Phishing scams
  • Celebrity pornography
  • Reputation smearing
  • Election manipulation
  • Automated disinformation attacks
  • Identity theft
  • Financial fraud

As noted previously, not all deepfakes are created for bad purposes. Take this one, for example. This campaign’s message is that WEtv has all 456 episodes of Law & Order.

Real or Not?

Which Face Is Real?: Learn to Spot Fake Faces at a Glance suggests that you look for clues like these to identify deepfakes: inconsistencies in the background; eyeglasses that aren’t symmetrical; unrealistic hair; odd/extra teeth; and other characteristics. See them all here

Other experts offer a long list of things to look for when identifying deepfakes. Here is a checklist of things to look for.

What about using technology to detect a deepfake? Using some tools that already exist such as a Google Reverse Image Search or Tineye can help find the original image. The Chrome extension RevEye Reverse Image Search does the same thing. etail/reveye-reverse-image-sear/keaaclcjhehbbapnphnmpiklalfhelgf?hl=en By finding the original image source, you can then determine if it has been manipulated, used to deceive, or been incorporated into synthetic media. 

For video try InVid. It is a ‘knowledge verification platform to detect emerging stories and assess the reliability of newsworthy video files and content spread via social media.’ It is aimed at journalists, but you can install the extension on Chrome or Firefox. It is free.

Both Facebook and Twitter have banned misinformation produced using AI, meaning “shallowfakes” are still allowed on the platforms. Their technology is evolving as deepfakes evolve, but it uses sophisticated algorithms to identify misleading videos. 

Perhaps the most important defense against deepfakes is to be skeptical about video you see on the internet or audio you hear over your phone or other means. Be sure what you see or hear  is legit and be careful about sending and sharing information until you are sure it is real. You don’t want to be part of the problem.

Let’s set the stage for the rest of this Thing. Try this quiz to see how well you can spot deepfakes. Spot the Deepfakes.


These videos offer some insight into deepfakes. 

  • Fake Videos of Real People — and How to Spot Them (7:15) | TED Talk


Learn more about  the impact and detection of deepfakes with these articles.


Put your new skills to a test by trying these activities. 

Attention check. This video is an attention check. Please watch and listen to this video and share how confident you are that it is fabricated. From MIT Media Lab. The Center for an Informed Public game Which Face is Real? is way harder than you might think. Tell us how you did in the Comments.

Minnesotans love to talk and share about weather, but can we trust images we see after a storm? Try a reverse image search using Google or Tineye using either of these photos. What did you find out? Add your results to the Comments.

Conversation Starters

Not all uses of deepfake technology are harmful. On balance, is deepfake technology a benefit or a detriment to society?

Which malicious actors are you most concerned about creating and spreading deep fakes in the immediate future, and with what goals?

Which library patrons/students (age, education…) do you think are most susceptible to which form of fake media: fake news (text) vs fake audio vs fake photos (spread with wrong context or doctored) vs fake video (manipulated and/or deep fake)?


Additional Resources

If you missed the Media Landscape presentation on Deepfakes with John Mack Freeman, you can view it here.

1 thought on “Thing 6”

  1. I think once we are aware of deepfakes/cheapfakes, we will hear about them everywhere! Case in point, the show FBI episode 3.14 was focused on deepfakes used to incriminate an innocent person, with devastating results (don’t @me–I know the problems with these kinds of show). Agent Bell even said at one point, “It’s a deepfake;” while they showed a screenshot & explained how they detected the deepfake–using reverse image search and the clues pointed out in the various articles above. Doubt I would have noticed this without the presentation & resources in this Thing! Where have you seen deepfakes or cheapfakes lately?

Leave a Reply