Bullhorn’s 92% AI Voice Interview Rating and Loyalty Gap

In its September 23, 2026 announcement, Boston-based recruitment software company Bullhorn said a survey of nearly 2,300 professionals who had worked with a staffing firm in the previous three years found that 92% of those interviewed by an AI voice agent rated it as good as or better than a live recruiter interview. That is a favorable rating from people who experienced a voice interview, rather than an approval rate for all job seekers.

The full candidate report says 57% of respondents had experienced a voice-agent interview and contrasts the 93% of candidates with a positive recruitment AI experience who would keep using their staffing firm with the 59% who said so after a negative experience. The loyalty comparison covers recruitment AI across the process, not voice interviews alone. It records what candidates said they would do, rather than whether they later returned.

Speed and relevant questions help explain the favorable rating

The findings point to two different ways automation can make an interaction feel useful. For chatbots, candidates valued an immediate response and a clearer sense of where they stood. For voice interviews, the accompanying accounts emphasize questions tied to a particular vacancy and follow-ups that let applicants explain skills a general screener might miss. The chatbot comments should not be read as a measured speed benefit of voice interviewing.

That distinction matters for recruiters choosing where to use an agent. A quick answer to a routine process question solves a different problem from a spoken assessment of technical or skilled-trade experience. Role-specific questions may give a candidate more room to show relevant knowledge, but the survey does not establish that an agent assesses that knowledge more accurately than a specialist recruiter.

The interview rating also concerns the conversation itself. An applicant can welcome a prompt, focused exchange while still needing a person to explain how it will affect the next decision. A favorable interview experience therefore says little on its own about the quality of the eventual shortlist or placement.

Negative experiences expose a wider process problem

The sharp loyalty split is a warning about the candidate journey around AI, not a measured failure rate for voice agents. People dissatisfied with recruitment AI were more likely to describe unclear next steps and a lack of help when they got stuck. They also expressed greater concern about being screened out unfairly or losing the chance to speak with a person. Those concerns can arise before or after an interview, as well as during one.

A candidate who cannot tell whether an automated conversation has finished, what happens next or whom to contact may lose confidence in the staffing firm even if the questions were relevant. These are plausible explanations for the association in the survey, not causes isolated by it. A disappointing hiring decision, weak recruiter communication and a frustrating AI interaction could occur together.

Human follow-up remains part of the service candidates are judging. A voice interview may move an initial conversation forward, but it cannot answer every question about an opportunity or a later decision. If a firm offers no clear way to reach a recruiter when the process breaks down, the apparent convenience of the interview may end at the point where the applicant needs help most.

The survey measures experience, not hiring quality

The respondents had recent experience with staffing firms across multiple regions, including North America, Europe and Asia-Pacific. That gives the findings relevance to agency recruitment, but it does not make them representative of every applicant or employer. The voice-interview rating comes from those who used an agent; the loyalty figures concern broader recruitment AI experiences and a different question.

A contemporaneous DHRMap brief notes the release but attributes the interview rating to Bullhorn, without independently testing the survey. Bullhorn sells recruitment software, and the published company pages do not disclose a separate voice-interview subgroup count, full question wording or a study validating hiring outcomes. Those omissions limit how precisely readers can interpret the rating and whether question wording influenced it.

Candidate satisfaction and hiring validity answer different questions. Testing whether voice interviews improve hiring decisions would require comparable applicants assessed under defined AI and human processes, followed by evidence on selection errors, job performance or other specified outcomes. This survey reports perceptions and intended loyalty. It cannot establish that AI identifies better hires, improves placement quality or treats applicant groups equally.

A design check for firms using voice screening

For recruiters, the practical response is to examine the parts of an automated interview that candidates can experience and question. The following are editorial design checks suggested by the findings, not remedies tested in the survey:

Role-specific questions: Can applicants explain relevant experience, and can follow-up questions capture context that a generic script would miss?
Process clarity: Before the interview, are applicants told they will speak with an AI agent, how the conversation will be used and when they can expect an update?
Escalation: Is there a visible route to a person if instructions are unclear, the technology fails or an applicant needs assistance?
Human follow-up: Is someone responsible for reviewing uncertain assessments, explaining the next stage and contacting the candidate as promised?

Those checks address the features candidates valued and the gaps dissatisfied candidates identified. As of the report’s release, the evidence supports a favorable account of voice-interview experience among users and a strong association between poor recruitment AI experiences and lower stated loyalty. Whether voice agents improve hiring decisions remains an open question requiring outcome data beyond this survey.