Will people talk to an AI interviewer? 99 calls say yes

We analysed 99 AI-led voice interviews from a live consulting project in Belgium, held in Dutch and French. The interviews were rated 8 out of 10 on average, the ratings held across every age group, and 86% of those who started stayed until the end.

Key takeaways

  • 8.0/10Average rating of the AI interview
  • 71%Rated their interview 8 or higher
  • 86%Of those who started stayed to the end
  • 8.0 vs 8.2Rated alike across every age group
  • 1 setupIn English, experienced well in Dutch and French

About the study

A consultancy used Cartorga to run AI-led voice interviews across Belgium as part of a live client project. Each participant was interviewed in their own language, Dutch or French, by an autonomous AI interviewer, and rated the interview from 1 to 10 at the end.

The interviews were ordinary phone calls rather than web calls. Phone lines add latency and make a conversation feel less natural, so these are harder conditions than a web interview.

Called
99
Started
84
Finished
72
Ages
23 to 73
Average age
46

Will people talk to an AI at all?

8.0/10Average rating, from 62 participants

Yes. Almost three quarters (71%) rated the interview 8 or higher, and about one in three (32%) gave a 9 or 10. The most common answer was 8.

“It is still a bit strange for now to talk with an AI, but I found the exchanges quite fluid.”Rating 8/10, translated from French
Exhibit 1.  Most participants rate it 8 or higher 01020Rating 1: 0 participants1Rating 2: 1 participants12Rating 3: 0 participants3Rating 4: 0 participants4Rating 5: 2 participants25Rating 6: 0 participants6Rating 7: 15 participants157Rating 8: 24 participants248Rating 9: 11 participants119Rating 10: 9 participants910
Number of participants per rating, 1 to 10. 62 ratings.

Does age matter?

NoAge makes no difference to the rating

Every age group rated the interview between 8.0 and 8.2 on average, and there is no relationship between age and rating. Participants aged 60 and over gave the highest average, 8.2 (from 6 ratings, a small group), and all 8 in that age group who started the interview finished it.

“I am really impressed by the ability of AI to conduct an interview like this one. It was very, very good.”Participant aged 59, rating 9/10, translated from French
Exhibit 2.  Every age group rates it about 8 0510All: 8.0Under 40: average 8.1 from 19 ratingsUnder 4019 ratings8.140-49: average 8.1 from 14 ratings40-4914 ratings8.150-59: average 8.0 from 21 ratings50-5921 ratings8.060 and over: average 8.2 from 6 ratings60 and over6 ratings, small group8.2
Average rating by age, out of 10. 60 ratings with a known age. No relationship between age and rating (Spearman ρ = −0.04, p = 0.79).

“You really have the feeling of being heard, of being listened to, because it is not just a dumb computer.”

Participant in the study, translated

Does it work in other languages?

7.9Dutch
8.2French
Average rating by language

Yes, from one English setup. The interviewer was set up once, in English, and ran in Dutch and French with no separate setup per language. Both groups rated it about the same: French-speaking participants slightly higher, a small difference that is not statistically significant (p = 0.34). In both, around 7 in 10 gave 8 or more.

Will people drop out halfway?

86%Of those who started reached the last question

Few do. Of 99 people called, 84 started the interview; the others could not be reached or chose not to take part. Of those 84, 72 finished: 83% in Dutch and 91% in French. Completion did not drop with age. In total, 62 participants rated the interview at the end.

Exhibit 3.  Most people who start, finish called: 9999 called85% of those called startedstarted: 8484 started86% of starters finishedreached the last question: 7272 reached the last question
All 99 people called, both languages.

Can it go deeper without tiring people?

7.6–8.4Average rating at every interview length

Yes. Short and long conversations were rated about the same. Whether an interview ran well under or well over the average length, its average rating stayed between 7.6 and 8.4, with no meaningful relationship between length and rating. Participants who kept it short were not dissatisfied, and those who talked at length did not tire of it.

That gives the interviewer room to follow up and probe where an answer is worth exploring, without hurting the experience.

Although we found no relationship between age and rating, we did find one between age and interview length. Participants under 40 talked 22% longer than the average interview, while those aged 40 to 59 were 11% to 13% below it, a statistically significant link. Their ratings were about the same.

Exhibit 4.  Longer interviews are rated just as well 0510All: 8.0More than 25% shorter than average: 7.8 from 13 ratingsMuch shorter13 ratings7.8Up to 25% shorter than average: 8.2 from 23 ratingsShorter23 ratings8.2Up to 25% longer than average: 8.4 from 13 ratingsLonger13 ratings8.4More than 25% longer than average: 7.6 from 11 ratingsMuch longer11 ratings7.6
Average rating, out of 10, by interview length compared with the average interview; “much” means more than 25%. 60 ratings. No meaningful relationship (Spearman ρ = 0.12, p = 0.37).

Why a few rated it lower

3 of 62 participants rated the interview below 7: one gave a 2 and two gave a 5.

The 2 came from a call with technical problems. The participant found the questions hard to understand, heard parts of the conversation in English and had choppy audio. We inspected the call and identified the technical problem behind it.

“A two. The questions were hard to understand. Things were often said in English. Yes, very choppy.”Rating 2/10, translated from Dutch

The two ratings of 5 gave little reason: one missed a human counterpart, the other saw it as a routine set of questions. Other critical comments focused on the voice’s regional accent and the timing between speakers, both settings that can be tuned per language. Nearly everyone who raised them still rated the interview 7 or higher.

What this means

  • Acceptance is not the bottleneck.People of every age and in both languages rate AI interviews around 8 out of 10.
  • One English setup is enough.Set up once in English, the interviewer ran in Dutch and French and was rated about equally in both.
  • There is room to go deeper.Longer conversations are rated as well as short ones, so the interviewer can follow up.

← All results