Real or AI? Spot the one real photo among five AI-generated fakes
Loading today's puzzle
Preparing images...
How to play
Full guide- 01
Look at all six
One is a real photograph. The other five were made by Google's Nano Banana image models.
- 02
Zoom into the details
Pinch, scroll or double-click. Fakes usually give themselves away up close.
- 03
Lock in one guess
Everyone gets the same puzzle each day. See how you compare with other players, and how the AI did.
AI Image Generation
How AI Generates Images & Why It Fails
Modern AI models like Google's Imagen and Stable Diffusion use a process called Diffusion. They do not draw like a human artist; instead, they start with random noise (static) and gradually refine it into a recognizable image based on patterns learned from billions of photos. While powerful, this statistical approach leads to the unique hallucinations you spot in the game.
Patterns vs. Understanding
The biggest limitation of current AI is that it has no concept of 3D space or anatomy. It knows what a hand looks like (texture, color, shape), but it does not understand that a hand must have five fingers and rigid bone structure. It is simply predicting pixels. This is why you often see melted objects, extra limbs, or architecture that does not make logical sense—the AI is mimicking the surface without understanding the structure beneath.
The Physics of Light
Light follows strict physical rules (reflection, refraction, shadows), but AI treats lighting as just another visual pattern. This leads to subtle tells: shadows falling in different directions, reflections that do not match the object, or pupils with different light sources. While a human artist instinctively adheres to physics (or breaks it intentionally), AI often creates subtle inconsistencies that our brains subconsciously recognize as wrong or uncanny.
Mastering these signs helps you spot deepfakes in the real world
Learn more about AI detectionWhy ImageGuesser Exists
We are living through a visual revolution. Generative AI models like Google's Imagen, Flux, and Stable Diffusion have reached a level of fidelity that rivals reality. For the first time in history, we cannot trust our eyes.
ImageGuesser is not just a daily puzzle; it is a digital literacy training ground. By gamifying the detection process, we help users build a mental library of “AI Artifacts”—the subtle glitches in physics, lighting, and anatomy that betray a synthetic image.
The Spotter’s Skillset
- Physics analysis: checking shadow consistency and reflection logic.
- Anatomical verification: counting fingers and teeth, and checking pupil symmetry.
- Contextual logic: identifying non-existent text or impossible architecture.
How We Generate Puzzles
Our daily challenges are curated using a strict ethical framework. Synthetic images are generated with Google's current Nano Banana image models (Gemini), so you are training against the state of the art. After you answer, each puzzle shows exactly which model made its fakes.
- Real photography
- Sourced from verified photographers on Unsplash with full attribution.
- Synthetic media
- Generated with Google's Nano Banana models - Nano Banana 2 and Nano Banana Pro today, the original Nano Banana in older puzzles.
- Review process
- Every puzzle is manually reviewed to ensure no deepfakes of real people are used.
Get in Touch
Have curious questions, found a bug, or want to discuss business? We’d love to hear from you!
