It’s becoming more and more difficult to distinguish between real images and fake ones online. And most people want to know, particularly when it’s not obvious. Video game images are frequently posted as actual war images and only those who played the games were able to spot them as the fakes they are. With the surging popularity of artificial intelligence image generators like Midjourney and DALL-E, creating fake images is simple: A few text prompts and you can create realistic pictures that often hold up well to scrutiny.

Thankfully, you soon won’t have to rely on your own powers of detection. Meta, the parent company of Facebook, Instagram and Threads (its X competitor launched last year), announced it will begin labeling third-party AI-generated images on all three platforms. Meta already flags images created by its own products with a visible watermark, and now it is working on ways to identify invisible markers in images from OpenAI, Google, Midjourney, Adobe, Shutterstock and others. While the company has not finalized the design of the identification, its post showed a label with “AI Info” under the account name.

Meta said it’s working with these other platforms to develop a unique marker that cannot be edited and would be standard across the industry. This way, Meta will be able to easily flag all participating companies’ images as containing AI-generated content. It said it will start applying labels in all languages supported by each app over the coming months.

Meta said the timing is important because “a number of important elections are taking place around the world,” which of course includes the U.S. presidential election this fall. And while doctored photos are important to identify, so are videos and accompanying audio. Unfortunately, the same technology cannot be applied – yet.

Watermarks or identifiers are rarely included in tools that generate audio and video. Meta said the industry is working toward this capability, but in the meantime, it will add a feature for people to disclose when they share AI-generated video or audio so Meta can add a label to it. When content is realistic, the user will be required to make the disclosure and Meta said it may apply penalties if they fail to do so.

Failing a label, social media users should be wary of content, especially if it’s inflammatory. Look at the source of the content to help determine whether or not the poster is trustworthy and read the comments that may expose possible AI-generated content. Take a closer look at the video itself and look for people or other things that look unnatural, similar to looking for signs of a Photoshopped image. Always cross-check with reliable news sources to see if the story is being covered and if the coverage is the same.

There are two sides to AI-generated content: Yes, it can be used to deceive, but it can also be used in a creative way to create art or just have fun. If you’re interested in trying your hand at AI-generated images, Meta offers two easy ways to get started.

In December, Meta opened its AI image generator tool called Imagine with Meta AI. You can use the generator for free by signing up on the website with an email address. You type in a text prompt and then Imagine will create four 1,280-by-1,280-pixel images that you can download.

It’s not magic. Like with any app built from a large language model such as ChatGPT, Imagine was trained using 1.1 billion images scraped from Facebook and Instagram posts. There are safeguards in place to ensure the images are family friendly, so the generator will refuse prompts that contain violent or sexual language. It also will reject famous people’s names. All images generated by Imagine are watermarked, making them easily identifiable as AI-generated pictures.

Meta also has built an AI-powered image enhancer into Instagram Stories that allows users to change their backgrounds in the images posted to Stories (as opposed to posting to the feed). On the Stories camera (swipe right from the home screen), look for the background icon next to the “add text” button. Tap it and enter a prompt describing the background you’d like in your photo. Like with Imagine images, these also will be identified as AI content.

Leslie Meredith has been writing about technology for more than a decade. As a mom of four, value, usefulness and online safety take priority. Have a question? Email Leslie at asklesliemeredith@gmail.com.

Newsletter