Do Employers Actually Call References? What They Ask in 2026
By
Liz Fujiwara
•
Mar 4, 2026
If you are an AI engineer, ML researcher, or infra specialist navigating the 2026 job market, you may wonder whether employers actually call references or if that is just an antiquated formality.
The short answer is yes; most serious employers still call references, especially for high-stakes technical roles, but the process has evolved to blend human judgment with AI-assisted screening for faster, more informed hiring decisions.
This article explains what to expect, what questions hiring managers typically ask, and how platforms like Fonzi are reshaping the reference check experience for technical candidates.
Key Takeaways
Most employers still call references in 2026, especially for AI, ML, and infra roles with high responsibility or compensation, and being asked for references usually indicates you are a front-runner.
Hiring managers use semi-structured questions focused on project impact, collaboration, reliability, and rehire likelihood, while AI tools help summarize notes and flag inconsistencies, but humans make the final decisions.
Curated platforms like Fonzi pre-validate candidates and references, reducing random reference calls and letting strong preparation turn reference checks into a competitive advantage.
Do Employers Actually Call References in 2026?
Let’s address the question directly: yes, many employers still call references, especially for roles with significant production responsibility or compensation above the market median.
For senior AI/ML roles, staff engineers, principal engineers, and founding engineers, structured reference calls are now standard practice. Smaller AI startups rely heavily on direct reference calls because they cannot afford hiring mistakes. Meanwhile, some FAANG+ companies and large unicorns may substitute internal transfer histories or detailed project documentation for external references when hiring from known networks.
Remote-first AI companies in 2026 are particularly diligent about reference verification. Without the ability to observe candidates in-office before hiring, they depend on references to validate aspects that interviews cannot fully reveal, such as how someone delivers under pressure or navigates ambiguous research requirements.
When in the Hiring Process Do Companies Call References?
Reference checks don’t happen at random. They’re positioned strategically in the hiring process, usually after the final interview loop when the team has identified a clear front-runner.
Here’s the typical timeline for technical roles:
Application submission
Recruiter screen
Technical screens (coding, system design, ML fundamentals)
On-site or virtual interview loop
Team debrief and decision
Reference checks
Verbal offer
Background check and security verification
Written offer
Being asked for professional references is a positive signal. Treat it as confirmation that you’re in the final evaluation stage, not as a mere formality or checkbox.
Some early-stage AI startups with fewer than 40 employees may call references right after a strong technical on-site to decide between two or three finalists. On Fonzi, companies often request references after Match Day once they’ve shortlisted one to three engineers per role, making the timing clearer and less stressful for everyone involved.
How Often Do Employers Call All Your References?
Employers rarely call every reference you list. Typically, they pick two to three references that best match the role and seniority level.
Many hiring managers start with one high-signal reference, and only contact others if the feedback is incomplete or ambiguous. This means quality matters far more than quantity.
Prioritize references who can speak to recent, high-impact work:
Launching LLM features
Scaling GPU infra for production workloads
Leading applied research projects that shipped
Managing incident response for critical systems
On curated platforms like Fonzi, companies come in with clearer role definitions. They ask for fewer but more targeted references, reducing noise for everyone and helping candidates avoid the awkward experience of over-notifying their former colleagues.
What Do Hiring Managers Actually Ask During Reference Calls?
Reference calls in 2026 have become more structured, especially for technical roles. Gone are the days of vague “What was it like working with them?” questions.
Modern managers focus on impact, collaboration, reliability, and growth potential. Many companies now use semi-standardized question sets to reduce bias and legal risk while still leaving room for open-ended feedback that provides valuable insight into a candidate’s real-world performance.
For AI and ML roles, hiring managers often ask about specific skill areas and project outcomes:
Offline metrics achieved on model deployments
Latency and throughput improvements delivered
Cost optimizations on compute infrastructure
Incident history and how the candidate responded
Ability to work cross-functionally with product, security verification teams, or legal
Recruiters and managers avoid asking about protected characteristics. Questions focus strictly on performance, work style, and whether the reference would rehire the candidate.
Some employers now record reference calls or feed anonymized notes into an internal system. References should be prepared to be quoted internally, even if their exact words aren’t shared with the candidate.
Common Questions Asked in 2026 (Especially for AI/ML Roles)
Here are concrete examples of questions you can expect former managers or supervisors to receive:
Question | What the Hiring Manager Wants to Learn |
“Can you verify [candidate’s] dates of employment and title?” | Basic fact-checking and ensuring resume accuracy |
“What were [candidate’s] major accomplishments in your time working together?” | Validating impact claims and depth of ownership |
“How did they handle ambiguous research requirements or shifting project scope?” | Assessing adaptability and problem-solving under uncertainty |
“Can you describe their collaboration style with product, infra, or other teams?” | Understanding cross-functional effectiveness |
“How did they navigate GPU budget constraints or infra incidents?” | Evaluating resource management and incident response |
“What are their biggest areas for growth?” | Identifying development needs and self-awareness |
“Would you rehire them if you had an open position?” | Getting a bottom-line recommendation signal |
“Is there anything else I should know about working with them?” | Surfacing unexpected insights or concerns |
Encourage your references to speak to these themes by sharing your target role description before they’re contacted. This helps them connect their answers clearly to required competencies.
Sample Reference Call Structure
A typical reference call lasts 15 to 25 minutes and follows a predictable flow:
Quick intro and context (2–3 minutes): The hiring manager explains the role and why they’re calling.
Basic fact verification (2–3 minutes): Confirming dates, titles, and reporting relationships.
Performance and behavior questions (10–15 minutes): Five to eight questions covering technical depth, work ethic, collaboration, and specific projects.
Closing questions (3–5 minutes): “Would you rehire?” and “Anything else I should know?”
Some companies follow up with a short email questionnaire to capture structured ratings, such as 1 to 5 scales for ownership, communication, and technical depth, which complement the call and help determine final decisions.
The tone is professional but conversational. Managers are not trying to catch you; they are looking to confirm patterns they already observed in interviews.
How AI Is Changing Reference Checks (Without Replacing People)
AI has entered the reference check process, but not in the way many candidates fear. Companies use AI to summarize reference notes, flag inconsistencies across multiple references, and suggest follow-up questions. The final decision, however, remains with human recruiters and hiring managers.
Responsible teams avoid feeding sensitive PII or full call recordings into third-party LLMs. Instead, they use internal tools or redacted summaries to protect privacy while still benefiting from AI-assisted analysis.
AI can also reduce bias in reference checks. By normalizing language and identifying when references systematically underrate certain groups, these tools prompt managers to focus on objective signals, such as shipped projects, incident history, and concrete metrics, rather than subjective impressions.
For candidates, this means your actual work product matters more than ever. The days of a single reference charming their way through a call are numbered when AI can cross-reference patterns across multiple perspectives.
How Fonzi Uses AI Differently in the Hiring Process
Fonzi is a curated marketplace built specifically for AI engineers, ML researchers, infrastructure engineers, and LLM specialists.
Fonzi uses AI to parse profiles, past work, and preferences to suggest high-signal matches to companies. This reduces the need for broad, generic reference checks because companies already come to conversations with validated information about a candidate’s background.
Importantly, Fonzi’s approach remains human-centered. Recruiters and hiring managers still meet candidates, run technical interviews, and interpret references with full context. AI helps them focus on people and does not replace the people-centric evaluation that matters most.
Fonzi’s Match Day: High-Signal Hiring Without the Noise
Match Day is a recurring event where vetted AI and infrastructure talent on Fonzi are introduced simultaneously to a curated group of hiring companies actively searching for their exact skills.
By the time Match Day happens, Fonzi has already collected key information, including skills, project history, preferences, and reference-ready context. Companies enter conversations with more trust and less guesswork, meaning the candidate pool they engage with has already been pre-qualified.
Benefits for candidates include:
Fewer random recruiter pings cluttering your inbox
More relevant roles aligned with your actual qualifications
Faster movement from first intro to final interview and references
Direct access to CTOs, heads of AI, or research leads instead of generic HR screens
Match Day is especially valuable for senior AI/ML professionals who want substantive conversations with decision-makers, not endless loops with junior recruiters who cannot answer technical questions.
How Match Day Differs From Traditional Recruiting
Traditional recruiting often involves scattered, months-long processes with cold outreach, ghosting, and unclear timelines. Match Day flips this model.
Instead of responding to malicious bots disguised as recruiter messages or wading through irrelevant job postings, candidates get several high-intent conversations in a narrow time window with companies actively hiring for their role.
How to Prepare Your References as a Technical Candidate
Choosing and preparing your references can make the difference between a lukewarm endorsement and a reference that actively strengthens your candidacy.
Select the right people:
Choose references who have seen you work on systems relevant to your target role
Prior managers on model deployment teams are ideal for ML roles
Tech leads from infra migrations speak well for platform engineering positions
Research leads or paper co-authors validate research contributions
Maintain an updated reference list that includes:
Name and current role
Your relationship to them
Dates you worked together
Two to three shared projects you want highlighted
Whether they’re better suited for research-heavy or production-heavy role discussions
Brief your references before they’re contacted:
Share the job description so they understand what the employer asks about
Remind them of specific projects where you demonstrated relevant skills
Explain your current narrative and career direction
Give them context on why you’re pursuing a new role
Align your stories:
Make sure the way you describe a 2026 project matches what your reference will share. Inconsistencies raise red flags.
Avoid name-dropping without substance:
Listing a “big name” reference who barely worked with you backfires. Depth of collaboration matters more than title for a credible, detailed reference call that can provide valuable insight.
Handling Potentially Risky or Neutral References
Not every past job produced a glowing reference relationship. Here’s how to navigate challenging situations.
Many former employers now have policies limiting managers to confirming employment dates and titles only. Don’t panic if a past supervisor can’t give qualitative feedback due to the company's culture around legal risk.
Alternative references to consider:
Tech leads who worked alongside you daily
Staff engineers who reviewed your code or designs
Cross-functional partners like PMs, data scientists, or SREs who saw your real contributions
If you have a known bad reference:
Be proactive with your recruiter; briefly explain context without being defensive
Provide stronger, more recent references to counterbalance older issues
Focus energy on demonstrating current capabilities through interviews and portfolio work
What Candidates Should Expect Companies to Ask References (With Comparison Table)
Reference questions in 2026 vary by role type, but patterns are predictable enough that you can prepare your references effectively. Understanding what employers want to validate helps you prime references to speak directly to those themes.
Role Type | Common Question Themes | What the Employer Wants to Learn |
AI/ML Engineer | Shipping models to production; tradeoffs between quality and latency; collaboration with infra and product; handling production incidents | Ability to deliver end-to-end, not just prototype; reliability under production pressure; cross-functional effectiveness |
ML Researcher | Experimentation cadence and rigor; balance between papers and product impact; collaboration with engineering teams; contribution to roadmap or IP | Research depth that translates to business value; ability to work within product constraints; long-term strategic thinking |
Infra/Platform Engineer | System reliability and scalability (GPUs, vector DBs); cost optimization; internal customer support; incident response leadership | Ability to keep critical systems running; resource efficiency; responsiveness to internal stakeholders |
Share your target role description with your references so they can emphasize the themes most relevant to the position you’re pursuing. A former supervisor who knows you’re targeting an ML researcher role will frame your work differently than if you were applying for a production engineering position.
Conclusion
Employers do still contact references in 2026, especially for AI and infrastructure roles where mistakes are costly and team dynamics matter greatly. Reference checks are not designed to catch you in a lie; they are about confirming patterns such as technical depth, ownership, and how you contribute to a team.
The right person with prepared references can turn this step into positive feedback that reinforces everything the interview revealed. Your former managers and colleagues become advocates who validate your resume claims and speak to aspects of your work that interviews cannot fully capture.
Ready to skip the noise and connect directly with top AI companies? Apply to join Fonzi’s talent pool, experience Match Day, and get in front of hiring managers who value your expertise and run reference checks the right way.
FAQ
Do employers actually call references, or is it just a formality?
At what stage of the hiring process do companies usually check references?
What questions do hiring managers typically ask during a reference call?
Can a bad reference cost me a job offer, and would I ever find out?
How many references do employers usually call before making a decision?
