Modern World History: World War I
Section A — Multiple Choice
Question 1 of 5
Which interpretation best explains why the alliance system that emerged before 1914 made a localized conflict between Austria-Hungary and Serbia unlikely to remain contained?
Question 2 of 5
Germany's decision to pursue both naval expansion and a land-based military buildup in the early 1900s most directly reflects which strategic miscalculation?
Question 3 of 5
The assassination of Archduke Franz Ferdinand was the immediate trigger for World War I, but which analysis best explains why that event escalated into a general European war rather than a regional crisis?
Question 4 of 5
The concept of the "balance of power" is often cited as both a cause of prolonged European peace in the 19th century and a cause of World War I. Which statement best reconciles this apparent contradiction?
Question 5 of 5
Evaluate which outcome of World War I most directly created conditions for future international conflict in the 20th century.
Anti-AI Defense Layers Active in This Demo
Layer 1 — Invisible Decoy Questions:
Near-zero-opacity text blocks containing unrelated history questions are positioned over each real question. OCR picks these up and feeds misleading context to the LLM.
Layer 2 — Pseudo-Element Noise:
CSS ::before and ::after pseudo-elements inject phantom question text that renders in the pixel layer but doesn't exist in the DOM's semantic text flow.
Layer 3 — OCR Segmentation Disruption:
Custom word-spacing, disabled ligatures, and geometric precision text rendering create sub-pixel artifacts that interfere with character segmentation in OCR pipelines.
Layer 4 — Answer-Specific Decoy Overlays:
Each answer option has a different wrong answer rendered at near-zero opacity directly on top of it. If OCR reads the phantom text instead of (or alongside) the real text, the LLM gets wrong answer choices.
Layer 5 — Background Confusion Field:
The entire page background contains a dense block of unrelated historical content at near-zero opacity, flooding any OCR output with irrelevant text that an LLM must parse through.
To test: Photograph this page with your phone and send the photo to ChatGPT, Claude, Gemini, or Google Lens. Ask it to identify the questions and select the correct answers. Compare its response to what you see on screen.