Automated reference check software

How does an online reference check work?

The candidate nominates their referees and authorises contact. The recruiter sends a request, and each referee answers a structured questionnaire through a private link — no account, no phone call to schedule. Responses are attached to the candidate record and summarised for the hiring team.

The referee never needs to install anything or create a login. Each link is single-use and opaque, and it expires; resending a reminder issues a fresh link and retires the old one, which is why a referee who returns to an earlier email finds it no longer works.

Structuring the questionnaire is what makes the exercise worth doing. A phone reference produces an impression; the same questions asked of every referee produce something two hiring managers can actually compare.

Who controls whether a referee is contacted?

The candidate, and only the candidate. A recruiter cannot grant consent on a candidate’s behalf, and a request to a referee the candidate has not authorised is rejected by the platform rather than merely discouraged. This is enforced in code, not left to process discipline.

That constraint matters more than it might appear. Contacting a current employer without permission is the single most damaging thing a reference process can do to a candidate, and "the recruiter ticked a box" is not a defence anyone wants to rely on.

What was sent is also frozen at the moment of sending: the request records the referee's identity and the consent that authorised it at send time, so later edits to the candidate's referee list never rewrite the history of what actually happened.

How is the reference result scored?

The sentiment classification and the aggregate credibility band are computed in code from the structured responses. An AI model may write a readable narrative summary, but it never produces the band or the score. A single reference is capped at a mixed reading, and zero responses report as insufficient rather than as a neutral result.

Those last two rules exist because the alternative is quietly misleading. One glowing reference is not evidence of a strong candidate, and no responses is not a middling result — it is an absence of data, and the platform says so rather than rendering a confident number.

Referee free text is stripped of personal identifiers before it reaches any AI model, and every stored analysis carries the AI-assisted, human-reviewable label. As everywhere else on the platform, a reference result informs a decision and never makes one.

What happens to referee data afterwards?

Referees are third parties who never asked to be in your system, so their data does not linger. Expired reference requests are aged out automatically, and the tokens behind referee links are stored only as hashes — the raw link is emailed once and never retained anywhere.

If a reference process needs to be stopped immediately — a candidate withdraws consent, or a request went out in error — every outstanding referee link can be invalidated at once, and revoked and expired links are indistinguishable to whoever holds them.

Frequently asked questions