The Study That Silicon Valley Can't Stop Talking About
This summer, researchers from the University of Chicago Booth School of Business and Erasmus University Rotterdam completed one of the largest natural field experiments ever conducted on AI in hiring, tracking over 70,000 real job applicants across 48 roles in healthcare, IT, and industrial sectors.
The results? AI voice agents didn't just match human recruiters - they exceeded them on every key metric.
What is AI Reference Checking?
AI reference checking uses voice agents to conduct automated reference calls, similar to the AI interviewing technology studied at University of Chicago. These systems ask structured questions, capture responses, and generate reports - completing in minutes what traditionally takes days. While the study focused on candidate interviews, the principles apply even more effectively to reference checking due to its structured nature. Learn more about how AI reference checking works.
The Numbers Don't Lie
The study revealed dramatic improvements when AI conducted interviews:
- 12% increase in job offers given to AI-interviewed candidates
- 18% more candidates showed up for their first day of work
- 17% higher retention after 30 days on the job
- 78% of candidates chose AI when given the option
These aren't lab results. These are real people, applying for real jobs at Fortune 500 companies.
Why AI Won (And Why It Matters for References)
The AI agent, named Anna, was trained to respond in real-time to a range of possible interview responses - a complex task requiring natural language understanding, contextual awareness, and dynamic conversation management.
Why Reference Checking Is Even Better Suited for AI
The AI agent was trained to conduct dynamic, real-time conversations—managing complex interview flows, follow-up questions, and nuanced candidate responses. This is genuinely difficult AI work.
Reference checks, by comparison, are remarkably straightforward:
| Task | Interview Complexity | Reference Complexity |
|---|---|---|
| Verify employment dates | N/A | Simple factual question |
| Assess cultural fit | Complex behavioral analysis | "How did they work with others?" |
| Evaluate skills | Multi-part technical assessment | "Rate their performance 1-10" |
| Predict success | Complex judgment call | "Would you rehire them?" |
If AI excels at complex interview processes, it should be exceptionally effective at simpler, more structured reference checks.
The Surprising Truth: People PREFER Talking to AI
When given the option to interview with the AI hiring agent, a staggering 78% of applicants opted for Anna over a human recruiter.
Why? Many candidates found the AI-driven interview less intimidating and more efficient.
For references - who are doing you a favor by taking your call - this preference is even stronger:
- No small talk required
- Get it done in 5 minutes
- No judgment or awkwardness
- Available 24/7 at their convenience
The "But What About the Human Touch?" Question
Dr. Brian Jabarian explained:
"It's not yet possible to delegate the more nuanced, in-depth evaluation of candidates entirely to AI. We will need human intervention to oversee and review the performance of AI recruiters."
This is precisely why Virvell handles references (structured data gathering) while recruiters handle relationships (nuanced human judgment).
For roles with more technical or routine tasks, incorporating real-time evaluation into the interview could truncate the long and cumbersome interview process. Reference checks are the definition of routine tasks.
What This Means for Recruitment Agencies
The research demonstrates that AI doesn't just automate processes—it actually improves outcomes:
- Better data quality: AI captured more job-relevant information in standardized formats
- Reduced bias: Consistent questioning eliminates interviewer variability
- Improved efficiency: Faster process without sacrificing (and actually improving) quality
- Enhanced candidate experience: Most candidates actually prefer the AI interaction
Applied to reference checking:
- Complete all references in hours instead of days
- Eliminate phone tag and scheduling delays
- Submit candidates while competitors are still calling references
- Win more placements through speed and thoroughness
See our pricing plans to understand the ROI of automated reference checking for your agency.
The Technical Reality Check
The study found some limitations:
- 3.2% of candidates ended interviews due to "AI aversion"
- 8% experienced technical system failures
- But 89% completed successfully even with first-generation technology
- Continuous improvement as systems learn
- Reference checking should see even better results due to its simpler, more structured nature
The Future Is Already Here
As of late last year, 93% of Fortune 500 Chief Human Resource Officers surveyed by Gallup said their organization had incorporated AI tools and technologies to enhance business operations.
The question isn't whether to adopt AI for reference checking.
The question is whether you'll do it before your competitors.
Implementation: The Virvell Advantage
Following the same principles demonstrated in the University of Chicago study:
- AI conducts structured conversations
- Captures standardized data
- Humans review and make decisions
- Continuous improvement through feedback
The difference? We've optimized specifically for reference checking - a task even more suited to AI than the interviews studied.
Frequently Asked Questions
Is the University of Chicago study specifically about reference checking?
No, the study examined AI job interviews with 70,000 candidates. However, the findings strongly support using AI for reference checking, which is even more structured and straightforward than interviews.
How accurate is AI reference checking?
While the Chicago study showed 89% completion rates for complex interviews, reference checking should see even better results due to its simpler, more structured nature.
How long does automated reference checking take?
AI reference checking completes in hours instead of days, eliminating the phone tag and scheduling delays that slow traditional methods.
Do references prefer talking to AI?
The study found 78% of candidates preferred AI interviews. For references doing a quick favor, this preference is likely even higher due to convenience and efficiency.
What's the ROI of AI reference checking?
Reference checks complete in hours instead of days. With faster turnaround, agencies can submit candidates while competitors are still waiting for callbacks, leading to more placements won.
Ready to implement AI reference checking?
Experience the power of AI reference checking firsthand with our free demo.
Study Citation: Jabarian, Brian and Henkel, Luca, Voice AI in Firms: A Natural Field Experiment on Automated Job Interviews (August 18, 2025). University of Chicago Booth School of Business and Erasmus University Rotterdam. Available at SSRN.