Do candidates actually like AI interviews? What 97% means

More often than anyone expects them to. 97% of candidates finish a Lantern interview satisfied with how it went, and the reasons they give are practical rather than enthusiastic: they could take it at a time that existed in their life, speak the language they think in, and hear back quickly.

A vendor quoting its own satisfaction number deserves scepticism, so this post spends most of its length on what candidates actually said rather than on the number. It also starts with the objection, because the objection is reasonable.

Why do candidates expect to hate an AI interview?

Because the phrase suggests being processed rather than assessed. The mental image is a webcam, a countdown, two retakes, something scoring your face, and no way to ask a clarifying question.

That expectation is generalised accurately from tools candidates have actually met. Plenty of what gets sold as AI hiring is a recorder with a scoring form attached, and a candidate who has sat through three of those is right to brace. The question is whether a live, adaptive conversation is the same object. It isn't, and the difference is structural rather than cosmetic — we set it out in AI interviewer vs. one-way video interview.

What changes their mind?

Usually the first follow-up. The interview asks a question, listens to the whole answer, and then asks something that only makes sense if it heard the answer — at which point the thing the candidate was dreading stops resembling the thing that's happening.

The same observation comes back in different vocabularies. Jason Chen, a candidate, put it as "felt like a real conversation, not a test." Adaptivity is the feature we build for hiring teams, because it's what produces signal; candidates experience it as being listened to, which turns out to be the same property viewed from the other chair.

Does it lower the pressure or raise it?

Lower, in most of what we hear. A live panel asks a candidate to think and perform simultaneously, and the performance half eats the thinking half.

Two other candidates described the same effect as a fairness gain rather than a comfort one. Tom B., a data analyst: "I actually got to show what I know instead of how nervous I was." Nicholas Crews, a candidate: "I actually got to show what I can do." Interview nerves are not a job-relevant trait for most roles, and a format that stops measuring them is measuring less noise.

How much of the satisfaction is just scheduling?

A large share, and we'd rather say so than dress it up as a breakthrough in conversational AI. The most common thing candidates tell us has nothing to do with the AI: it's that the interview happened when they were free.

For anyone employed, on shift, or caring for someone, a weekday-daytime screen means taking leave or lying to a current employer. Ashley James, a candidate, described the fix plainly: "I could do it on my own time — so much less stressful." Sofia M., a designer, named the part that usually goes unmeasured: "Rescheduling was one click. Compare that to emailing a recruiter five times." Drop-off in most funnels is a scheduling problem wearing the costume of disinterest.

What does interviewing in your own language change?

It changes what's being assessed. A candidate answering in a second language is spending part of their attention on translation, and any interviewer — human or not — is partly scoring the translation.

Amara O., in marketing, described the adjacent relief: "No awkward small talk, no being judged on my accent in the first five seconds." Lantern interviews in 40+ languages and dialects natively rather than in translation, on one rubric audited for language drift. The mechanics, including what happens when someone switches language mid-answer, are in multilingual AI interviews, and the coverage list is on the languages section.

Why does response speed matter this much?

Because the alternative candidates are comparing it against is silence. Being ghosted after an interview is the default experience in most markets, and it costs a candidate nothing to say so publicly.

David K., in sales: "Got feedback the same week. Most companies just ghost you." Marcello Raphael, a candidate: "Fast, fair, and I heard back quickly." Connor Woods: "Way better than waiting weeks for a callback." None of that is about the interview. It's about a funnel that no longer has a two-week hole in the middle of it, which is the same operational change described in what 70% faster time-to-fill actually looks like.

Common questions

Do candidates know they're being interviewed by an AI?

Yes. Transparency to the person being assessed is a core expectation of the high-risk regime we build against, and we design for candidates knowing rather than finding out. More on that in how we approach the EU AI Act.

What happens if a candidate refuses to do it?

It goes to the client's team. Lantern never rejects anyone, and declining the format isn't a scoring event — it's an escalation, handled by a human who decides what the alternative looks like.

Does satisfaction vary by role type?

We publish one number across 100K+ interviews rather than slicing it, because slicing a satisfaction metric until it flatters you is easy. The strongest reactions we hear come from shift workers and from candidates interviewing outside English. The honest recommendation is to test it on your own pipeline with your own candidates.

The fastest way to find out whether your candidates like it is to ask yours rather than ours. Run one against a live req.