AI Interview Tools Claim .35 Validity — Here's Why That's Not Good Enough

A major AI interview vendor cites a validity score of .35 in their own technical report and frames it as competitive. But competitive with what, exactly? IO psychology research shows structured behavioral interviews hit .48 to .51. That means some AI tools are selling you unstructured-interview-level accuracy at enterprise software prices. In this episode, we dig into what the independent research actually says about AI interview scoring — and it's a lot more complicated than the pitch decks suggest. From a University of Melbourne study that found every single tool they tested was biased against non-native English speakers, to SHRM data showing only 22% of employers even test their AI hiring tools for bias, there are real gaps between the marketing and the science. We also cover the regulatory wave closing in: EEOC employer liability, NYC Local Law 144 bias audit requirements, and the EU AI Act's high-risk classification for employment AI. Regulators are putting the burden squarely on employers, not vendors. If your company is using AI to score interviews — or thinking about it — this episode gives you five specific questions to ask before you sign anything.