Methodology

Sourcing and import

Test material is imported from source archives as images and/or text. During import, every item is individually inspected: question content, answer options, ordering, and completeness. We do not rely on filenames or assumptions — each item's content is verified visually before it enters the structured content model.

Review before publication

Items that cannot be verified with confidence are flagged NEEDS_REVIEW in the content system and documented rather than silently guessed. Flags remain attached to the content and are re-examined before any scoring change.

For our visual matrix IQ test, this process identified two near-duplicate items in the source archive and one item with an ambiguous solution; all three are documented on the test's landing page under limitations.

Scoring

Each test declares its own scoring configuration — correct-answer keys, band definitions, and (where applicable) estimate anchors. A deterministic scoring engine applies those rules server-side. The same inputs always produce the same result, and AI plays no role in calculating scores.

Estimates and norms

Where we translate raw scores into familiar scales (such as an IQ estimate), we interpolate between anchor points based on published norms for the test format, and we label the output as an approximation. Online self-administration differs from supervised testing, so estimates carry meaningful uncertainty and are presented as educational.

AI interpretation

The optional AI interpretation receives the calculated scores and the test's scoring rules — never your identity. Its output is validated against a strict structure, screened for diagnostic language, and always displayed beneath the deterministic result with a disclaimer.