AI or Real Quiz

For anyone judging questionable online images more carefully, the AI or Real Quiz offers a structured way to examine evidence instead of relying on instinct. As an AI or not quiz, it suits curious viewers, students, journalists, and people who share visual claims. The questionnaire assesses whether an image is AI-generated, edited, or camera-captured, while showing how you weigh clues, sources, and uncertainty.

What This Assessment Measures

Questions address how an image may be AI-generated, edited, recompressed, miscaptioned, or camera-captured. As an AI image quiz, it tests clues without treating any as proof.

The scale, AI or Real Image Knowledge (0-16), summarizes knowledge of AI-image clues, camera artifacts, provenance, and verification. It does not turn the exercise into a real or fake quiz that treats one clue as decisive. Its bands are Foundation (0-6), Developing (7-12), and Strong (13-16). A higher band reflects use of multiple evidence types and comfort with uncertainty, not proof of authentication skill.

Themes include:

• Visual anomalies prompt investigation rather than proving generation.
• Provenance and metadata show a file’s history, not every claim about its subject.
• Context and source tracing expose recycled images, misleading captions, or missing background.
• Uncertainty handling shows whether competing clues are weighed before a conclusion.

Methodology and Evidence Base

Each item presents a verification scenario with four choices, making this an AI vs real photo quiz about evidence rather than visual instinct. The respondent selects the most defensible interpretation, treating malformed hands, conflicting shadows, distorted text, missing metadata, and detector percentages as clues whose reliability depends on context. Judging whether a photo is AI-generated or camera-captured therefore requires evidence evaluation, not recognition of one visual signature.

The approach emphasizes corroboration: inspecting the best available file, comparing nearby details, tracing earlier appearances, checking provenance, and distinguishing compression artifacts from manipulation. The questionnaire displays a 0-16 score, but its item-weighting or calculation method and norm-based standardization are not specified.

The questionnaire is conceptually consistent with Cooke et al. (2025), whose findings support caution about unaided judgments of synthetic content, and with C2PA and NIST AI 100-4 on provenance and digital transparency; it is not an adaptation of those published works. These sources support combining signals rather than treating one detector, artifact, or credential as conclusive. Platform processing, missing provenance, misleading context, source quality, and current knowledge can affect answers, while the score reflects educational knowledge of evidence types and their limits.