Not trying to be inflammatory but. ive been quietly watching this space for two years and im less convinced now than i was in 2024 that any of this actually works at scale. the detector companies all market different headline accuracies, but theres no shared benchmark. the independent academic papers that test them tend to find they fall apart in the wild. and yet they’re getting deployed across universities, publishers, and even hr screening. discuss?
Not a scam, but oversold. the technical methods are real, they just dont generalize as well as marketers claim. the gap between lab performance and field performance is large. the deployment by universities is the bigger issue: its outpaced the evidence base by maybe 18 months.
Scam is harsh. but the marketing is irresponsible. ive seen vendors quote ‘99.8% accuracy’ in sales decks based on internal test sets that nobody outside their company has ever seen.
the industry will mature, but right now its the wild west. the better tools will publish their evaluation methodology and ride out the next 12-18 months. the worse ones will get shaken out when the first major lawsuit lands.
Yeah i think ‘oversold’ is fairer than ‘scam’. its just frustrating watching big institutions adopt this stuff with so little scrutiny