Theme 01
It feels like a real person.
The single most frequent unprompted positive theme. Candidates describe forgetting, mid-call, that they're not speaking with a human.
“Felt as if I was talking to a real person on the other side.”
Logistics role
A Classet Research Report · Issue Nº 01 · April 2026
Do candidates actually like AI interviews?
We asked 1,125 of them. 88% said yes. The honest watchouts are inside.
Download the reportFree. Work email required.
Classet makes Joy, an AI phone interviewer. We had every reason to spin our own data. We didn't. The honest watchouts are inside.
Classet · Classet.ai
Volume 01
Issue Nº 01 · Candidate Feedback
What candidates actually say when AI is on the other end of the call.
April 20261,125 candidates
1,125
Post-interview ratings
66
Organizations
9
Industries
48.4%
Leave written feedback
Section 03 · The Headline Numbers
What candidates actually rate AI interviews.
88% of candidates rate the experience positively. 64% pick Excellent — the top mark on a four-point scale. The negative share is 12%, and most of it isn't about the AI.
Chart 01 · Rating distribution
Four-point scale, 30-day window ending April 2026.
64%
24%
5%
7%
- Excellent64%· Top mark
- Good24%· Positive
- Fair5%· Neutral
- Poor7%· Negative
Positive · 88%Negative · 12%
“Best AI I've ever encountered. Felt human.”
Section 05 · The Qualitative Heart
Five themes that come up over and over.
Five themes show up in nearly every industry and both languages we sampled.
Theme 02
Less pressure than a human.
Candidates feel less judged and more able to think on an AI call. First-time interviewers and non-native English speakers say it most.
“I was more comfortable talking to the AI interviewer than I will be talking to a human.”
Healthcare role
Theme 03
Fast, simple, out of the way.
Candidates describe the experience as easy and fast. They value the absence of scheduling and knowing where they stand.
“Quick, informative, and responsive.”
Manufacturing role
Theme 04
Good even when something goes wrong.
When something does go wrong — dropped mic, lost connection — many candidates still open with praise for the AI. The underlying experience holds.
“It was good at the beginning, but the AI interviewer suddenly froze.”
Skilled trades role
Theme 05
Works across languages.
Spanish-language feedback is almost entirely positive. Candidates with limited English explicitly note that the AI helps them, not hinders.
“My English is not so good, but the AI helped much.”
Hospitality role
Section 06 · The Honest Watchouts
Where AI interviewing struggles. And what to do about it.
No candidate feedback dataset of this size is all green lights. The issues are real, but small in aggregate, and almost all of them are fixable on the employer side.
The full report includes five questions to ask any AI interviewing vendor before you sign.
- 6.1
Pacing and interruption
Candidates who pause to think, translate, or process can be cut off. There's a real accessibility dimension worth asking every vendor about.
- 6.2
The screen-timeout trap
Phone sleeps, mic stops, candidate doesn't notice. A pre-call instruction to keep screen awake fixes most of this.
- 6.3
Transcription accuracy
Names and uncommon words get mangled. Candidates are unnerved when they see their name misspelled in the recruiter view.
- 6.4
Echo on speakerphone
The AI hears its own voice and transcribes it as the candidate's answer. A pre-call headphone nudge solves it.
- 6.5
The small minority who prefer humans
About 1% explicitly want a human recruiter. Keep a clear, fast path for them. It's good candidate experience and good sense.
Section 07 · The Root-Cause Reveal
When candidates rate negatively, the AI itself is usually not the reason.
We classified every negative rating with written feedback. The majority trace to candidate-side infrastructure — a microphone that wasn't working, a network that dropped, a phone screen that timed out.
Chart 02 · Root cause of the 12% negative ratings
Each bar is a share of the 12% who rated negatively, not the full sample. Only one root cause is the AI itself.
38%
Disconnection / call crash
Non-AI cause
22%
Candidate mic / audio
Non-AI cause
20%
AI product issue
AI cause
12%
Candidate environment
Non-AI cause
8%
Unclassified / no written feedback
Data gap
Shares are of the 12% negative ratings, not the full 1,125 sample.
Worried AI interviewing will trigger complaints? Look at what candidates actually complain about. Most of it is infrastructure and instructions. Both of which any TA team can address.
The Free Download
Get the full 25-page report.
Plus the practitioner playbook, six objections answered with data, and the full methodology.
- Ten editorial sections, with charts and quotes
- The five practices that separate AI interviewing done well from done badly
- Six common objections, answered with the data
- Five questions to ask any AI vendor before you sign
- Full methodology, sample, and limitations
Download the report
Instant access. Work email required.
By submitting, you agree to our Privacy Policy. We'll never share your info.
In a candidate's own words
“It felt like I was talking to a real person. The AI made me feel comfortable and confident that I am making the right decision to apply for this job.”
Candidate · Hospitality role
See how Joy runs these interviews