TestGorilla Survey: Recruiters are Faster with AI, but Not More Confident They’re Finding the Right Candidates
A TestGorilla-commissioned survey of US recruiters reveals a telling paradox: AI is making the search for candidates faster — 72% of respondents said so — but it isn't making recruiters more certain they're choosing the right people. Roughly one-third saw no gain in hiring confidence, and 18% said their confidence actually fell.
Key facts
- 72% of surveyed US recruiters said AI accelerated their candidate search.
- ~33% (one-third) reported no improvement in confidence about finding the right candidates.
- 18% said their confidence in identifying the right hires actually declined with AI use.
- The survey was commissioned by TestGorilla, a skills-based hiring assessment platform.
- Survey population is US recruiters; sample size and methodology have not been publicly disclosed.
- Findings were distributed via Business Wire.
AI recruiting tools are delivering on their promise of speed — but not necessarily on quality of hire, according to a new survey by skills-assessment platform TestGorilla.
The study, fielded among US recruiters and released via Business Wire, found that 72% of respondents said AI had accelerated their candidate search process. That figure will come as little surprise to an industry that has rapidly integrated AI-powered sourcing, screening, and ranking tools over the past two years.
What is more striking is what happened to confidence. Approximately one-third of recruiters surveyed said AI brought no improvement in their certainty that they were identifying the right candidates. More notably, 18% reported that their confidence in finding the right hires actually declined after adopting AI-driven approaches.
Taken together, these findings point to an emerging gap between operational efficiency and decision quality — what might be called the 'speed-confidence paradox' in AI-assisted recruiting. Recruiters are processing more candidates, faster, but a meaningful share are not feeling better about their ultimate choices.
TestGorilla, which sells skills-based assessment products designed to improve hiring accuracy, has a clear commercial interest in surfacing this gap. Still, the directional finding aligns with concerns raised by talent acquisition professionals and researchers who argue that AI tools trained on historical hiring data can perpetuate bias, over-index on resume signals, or simply shift the bottleneck from sourcing to evaluation without improving the latter.
Full survey methodology, sample size, and fielding dates had not been disclosed in available materials at the time of publication. The survey covers US recruiters only, and findings may not reflect sentiment in markets where AI adoption in HR is at a different stage.
For HR technology vendors and practitioners alike, the data underscore a product design challenge: building AI that doesn't just surface candidates faster, but gives recruiters demonstrably better signal about fit, potential, and retention risk.
Verified vendor profiles for the companies named above — DHRMap wires every story to its directory.
Pre-employment testing platform offering a library of skills, cognitive and personality assessments to screen job candidates.
Paradox makes conversational recruiting AI (assistant 'Olivia') that automates candidate screening, interview scheduling and onboarding.
Reflect is French HR analytics software that centralises workforce, recruitment and HR data quality dashboards.
DHRMap tracks every HR-tech company over time — here's the rest of the coverage.
The survey crystallises a tension that is becoming central to the AI-in-HR debate: optimising for speed is not the same as optimising for outcomes. If AI tools make hiring faster but leave a significant share of recruiters less confident in their decisions, it raises urgent questions for product teams, HR leaders, and the vendors selling these solutions. For DHRMap readers — HR tech buyers, builders, and investors — this is a data-backed signal to scrutinise 'AI efficiency' claims and demand evidence on decision quality, not just throughput.
