REAL / FAKE / AI: can you tell what is real?
REAL / FAKE / AI is a free 15-minute game for teenagers aged 14 to 18 that teaches how to tell real photos, AI-edited photos and AI-generated images apart, and how to check any image before sharing it. Andrei's feed is full of viral posts: a flooded boulevard, a giveaway, a kitten in the rain, a school buried in snow. Some are real, some are real photos edited with AI and some never happened at all. Sort fourteen posts, find what was changed in three edited photos, learn how professionals verify an image and decide what to do when a deepfake targets a classmate. At the end you get a score out of 100 and a 60-second verification guide.
- About 15 minutes
- 5 steps, 14 posts, 2 videos
- Score out of 100
- No uploads, nothing sent online
About REAL / FAKE / AI: a game about AI images and deepfakes for teens
REAL / FAKE / AI is a free, interactive media-literacy game created by Nicholas Carstoiu (Nicholas Cârstoiu) as Module 04 of TRUST // NO TRUST, a digital resilience platform for teenagers. It teaches how to tell authentic photos, real photos edited with artificial intelligence and images or videos generated entirely with AI apart. You scroll through a fictional feed of viral posts, find what was changed in edited photos, put the steps of image verification in order and decide what to do in two real-life situations, one of them a deepfake of a classmate. Along the way it connects media literacy, AI literacy, misinformation awareness and the responsible use of generative AI.
What does it teach?
You learn that a real photo can look unbelievable and a fake can look completely ordinary, that edits are often tiny, and that the newest AI images rarely show the classic mistakes. Above all, you practise what works: checking the account, searching the news, running a reverse image search, looking for AI labels and Content Credentials, and labelling your own AI content honestly. The final step shows how to protect a classmate who is targeted by a deepfake.
Who is it for?
REAL / FAKE / AI is designed for teenagers aged 14 to 18 and takes about 15 minutes. It works on your own or as a classroom media-literacy activity, for example in a form-tutor lesson, and ends with a score out of 100. It also explains Article 50 of the EU AI Act, the transparency rule for AI-generated content and deepfakes that applies from 2 August 2026. Nothing is uploaded and no account is needed.
REAL / FAKE / AI follows TRAP, where you spot the moment a friendship turns into manipulation, and links back to SURVIVE, the simulated week online.
Photo credits: thank you to Kanenori and Stephanie Hanstein for two of the real photos. The AI images and AI edits were made with an OpenAI image model and the AI video with Kling AI, specifically for this exercise.
How to verify an image before sharing it (this also gives away the answers to step 3)
A convincing image is not proof that something happened. Professional fact-checkers do not rely on a single trick; they combine a few quick checks:
- Pause. If a post makes you scared, angry or rushed, slow down before you share it.
- Check the account. Look at how old it is, what else it posts and whether it is the original source or a copy of a copy.
- Look for independent confirmation. Something big will be reported by trusted news outlets or official pages, not only by one viral account.
- Search the image itself. A reverse image search (Google Lens, TinEye) shows where it appeared first and whether it is old or from somewhere else.
- Check labels and provenance. Look for AI marks and captions, and check Content Credentials when you have the original file. Missing credentials do not prove that an image is fake, and visual details alone cannot prove that it is real.
- Decide. Share it with context, report it or ignore it. If a deepfake targets someone, do not amplify it.
REAL / FAKE / AI, created by Nicholas Carstoiu, turns these checks into an interactive exercise.
Questions and answers
What is REAL / FAKE / AI?
REAL / FAKE / AI is the fourth module of TRUST // NO TRUST: a short game in which you sort a fictional feed of viral posts into real, edited and AI-generated images, then learn how to check any image before you share it.
Are the images real?
Some are. The feed mixes real photos, real photos that were edited with an AI tool and images generated entirely with AI. The accounts and captions are invented, the people in the AI images do not exist and every image carries a small "TRUST // NO TRUST · exercise" mark so it cannot be passed off as real.
How is the score calculated?
The five steps are worth 40, 20, 15, 10 and 15 points. In the feed you earn a full point for every correct answer and half a point when you see that an image is not real but mix up "edited" and "AI-generated". Everyone sees the same feed, so scores can be compared.
Can you tell an AI image just by looking at it?
Not reliably. The newest AI images often have none of the classic mistakes, such as wrong hands or blurred text, and a real photo can look unbelievable. Check the source, the context and whether anyone independent confirms it before you decide.
What are Content Credentials?
Content Credentials are provenance information, based on the open C2PA standard, that can be attached to a file to record how it was made and edited. Many AI tools now add them, and you can read them at contentcredentials.org/verify. They disappear from screenshots and re-saved copies, so missing credentials do not prove that an image is fake.
What should you do if someone shares a deepfake of a classmate?
Do not like, comment on or forward it. Tell your classmate you believe them, save the evidence (a screenshot with the link and the account name), report the post on the platform and tell a trusted adult or your form teacher. If the image or video is sexual, do not download or forward it at all: report it and tell an adult straight away.
