This service is built from a pre-compiled set of known AI-generated sites and their markers. These markers are stored and paired with weights to be used as reference patterns when analyzing user-submitted data.
How scores work: Scores are computed by counting found markers within each assessment. These are made in the 0-100 range, and are combined into a single total score that's normalized back into the reference set.
How weights work: Weights control influence per assessment, the larger weights acting as a positive multiplier.
The general formula is as follows.
final_total = (Σ(scorei × weighti)) / (Σ(100 × weighti))
How tiers work: Tiers are derived from the total score by fixed bands: weak (0-34.99), medium (35-64.99), strong (65-100). These cutoffs are tuned against the reference catalog so that each tier reflects how often similar scores appear in practice.
General factors considered: DOM structure, section layout, heading patterns, button and CTA copy, emoji usage, repetition, placeholder strings, builder tags, framework classes, bundle names, url shape, meta text, and other stored markers taken from the reference set.
Features missing? Feel free to contact us with your suggestions!