AI or Real Quiz
Questions: 16 · 10 minutes
1. A file has verifiable signed provenance recording camera capture followed by cropping and color adjustment. What does that evidence establish most directly?
Every visible event in the scene happened exactly as presented
The image contains no misleading framing or omitted context
A documented source and editing history for the file, not the truth of every claim about its subject
No generative tool could have been used anywhere in the workflow
2. A photograph appears to show an impossible giant creature standing beside a building. Which conclusion is most defensible from visual implausibility alone?
The scene may involve a model, staging, compositing, perspective, or generation, so implausibility alone does not identify the source
The creature must have existed because buildings provide scale
The file is necessarily a fully AI-generated image
The camera metadata will necessarily reveal what the creature is
3. A pictured hand appears to have six fingers. What is the most accurate interpretation?
It proves the entire image was generated by AI
It guarantees that the hand was manually pasted into the scene
It specifically indicates a defective camera lens
It is a useful anomaly to investigate, but it is not proof by itself
4. In a storefront image, several sign letters merge, repeat, or change shape within the same word. What should this pattern suggest?
The image must be real because compression only affects authentic photographs
Generation or editing is plausible, so the original file and wider context should be checked
The sign is necessarily written in an unfamiliar language
The photo was certainly taken in low light with a long exposure
5. A mirror reflection shows a person’s raised arm in a clearly different position, and the viewing angle does not explain the mismatch. What is the best response?
Declare the image AI-generated without considering compositing or other edits
Treat it as a local inconsistency and inspect the original, nearby details, and possible editing history
Ignore it because reflections never provide useful visual evidence
Assume the mirror itself changed the timing of the reflection
6. You find some clues suggesting generation and others supporting camera capture. What is the most reliable way to report your judgment?
Choose “real” because convincing lighting outweighs source information
Choose “AI” because any suspicious clue should override all other evidence
State the uncertainty, describe the competing evidence, and seek better source or provenance information
Average the clues into an exact probability without considering their reliability
7. A news photographer provides a file claimed to be an original camera photo. Which additional evidence would most strongly support that claim?
The original RAW file, adjacent shots from the same sequence, and mutually consistent capture metadata
A high-resolution JPEG with unusually sharp facial details
A visible news-agency watermark in one corner
Thousands of shares from accounts calling the photo genuine
8. A reverse-image search finds the same photograph on a reputable website from 2018, although a current post claims it shows an event from this week. What is the strongest conclusion?
The photograph must have been generated in 2018
The image predates the current claim, so the new caption is misleading even if the image itself is camera-captured
The current post is accurate because old images can resemble new events
Reverse-image search establishes every edit made to the file
9. An image’s metadata names photo-editing software in the “Software” field. What does that fact demonstrate?
The file passed through that software, but its complete origin and degree of editing remain uncertain
Every object in the scene was manually altered
The image was never captured with a camera
The whole scene was generated from a text prompt
10. Two people standing on the same flat ground under one visible sun cast shadows in sharply opposing directions. How should this clue be handled?
It should be dismissed because lighting can never reveal manipulation
It proves the people were generated rather than separately photographed
It confirms that two suns were outside the camera frame
It is a meaningful lighting inconsistency that warrants checking for generation, compositing, or another explanation
11. Which workflow gives the most dependable assessment of whether a suspicious online image is AI-generated, edited, or camera-captured?
Inspect the hands and stop if every finger looks normal
Examine the best available file, check internal consistency, trace earlier sources, review provenance, and express remaining uncertainty
Judge the image from the caption and the number of reposts
Trust whichever automated detector gives the most confident percentage
12. Which statement about hands and faces in modern AI-generated images is most accurate?
Perfectly symmetrical facial features prove that a portrait is authentic
Every generated portrait contains at least one malformed finger
They can appear anatomically convincing, so the absence of obvious defects does not prove camera capture
A camera can never produce odd-looking anatomy through motion, perspective, or occlusion
13. An automated detector reports that an uploaded image is “92% AI.” How should that result be used?
As conclusive proof because a percentage above 90 is definitive
As evidence that the image cannot contain any camera-captured material
As proof that exactly 92% of the image’s pixels were generated
As one imperfect signal whose reliability can vary with the detector, image type, and processing
14. Why is distorted lettering in a tiny, heavily compressed image weaker evidence than distorted lettering in the full-resolution original?
Downsampling and compression can deform genuine letters and create misleading shapes
Only camera photos lose detail when resized
AI-generated text automatically corrects itself at low resolution
Metadata becomes visible inside letters when an image is compressed
15. An image downloaded from a social platform contains no camera metadata. What can you safely conclude?
It was generated by AI because cameras always preserve metadata online
It was captured on film and later scanned
Its creator intentionally removed evidence of manipulation
Very little; platforms, screenshots, exports, and edits can all remove metadata
16. A low-quality image shows square blocks and bright halos around nearly every high-contrast edge, including buildings, people, and text. What most directly explains that repeated pattern?
Incorrect scene geometry created during generation
A signed record of camera ownership
Lossy compression or repeated saving of the image
Natural atmospheric haze at the location