Detection Tools Cover Only 45% of Known Pattern Types
Among 8 dark pattern detection tools evaluated against a comprehensive 68-type taxonomy, collective coverage reached only 31 types, a 45.5% coverage rate.
The best-performing individual tool (AGM) detected 15 types, achieving just 22% coverage. 37 dark pattern types (54.5%) remained undetectable by any available tool.
This gap persists despite the tools using advanced techniques including computer vision, NLP, transformers, and large language models.
Related: 02-atom—taxonomy-granularity-affects-detection, 04-atom—dataset-coverage-limits-detection, 07-molecule—taxonomy-to-detection-gap