A quarter of all interviews conducted through one voice-AI recruitment platform now happen between ten at night and two in the morning. For manufacturing employers it's over a third. Nobody has scheduled these. There is no interviewer sitting there yawning at half past midnight, because there is no interviewer. There is a candidate, a laptop, and a model waiting patiently to be talked at.
The industry frames this as flexibility. It is worth asking flexibility for whom.
Availability is not the same as choice
The defence writes itself, and it isn't entirely wrong. Parents get the children down and finally have an hour. Shift workers finish at ten and can't take a call at eleven in the morning because they're asleep. For these people, a system that never closes is a genuine improvement on one that only operates during the hours they are contractually obliged to be elsewhere.
But there's a sleight of hand in it. The problem being solved — that the working day is structured so that jobseekers cannot job-seek during it — is a problem employers created and could have solved differently. Offering interview slots in the evening with an actual human would also fix it. Instead the fix is to remove the human entirely and let the machine absorb the inconvenience, which happens to cost nothing and happens to scale infinitely.
What looks like accommodation is really the extension of the labour market into the last unclaimed hours of the day. The night was, until fairly recently, the one time you were reliably not being assessed.
Screaming into the void
The deeper complaint isn't the hour. It's the opacity. Candidates don't know what the system is measuring. Was it the pause? The regional accent? The failure to deploy some magic phrase the model was trained to reward? One person described the experience as shouting into nothing, which is about right — you perform, sincerely, to something that will not react, and then silence.
And people are voting with their feet. Greenhouse's 2026 survey of nearly three thousand active jobseekers found 38% had already withdrawn from a hiring process because it involved an AI interview, with another 12% saying they would if it were compulsory. Crucially, the same research found only around one in five actually want less AI in hiring. They want disclosure, an explanation of what's being assessed, and a human somewhere in the decision chain.
That's not technophobia. That's a fairly precise account of procedural fairness, articulated by people who have been on the wrong end of its absence.
The asymmetry nobody mentions
Employers adopted AI screening because applicants started using AI to write applications, flooding the funnel. Applicants started using AI because the funnel was already a lottery. Each side automates in response to the other's automation, and the result is a system where a machine writes a letter that a machine reads, and somewhere in the middle a person's livelihood is decided by two statistical processes talking past each other.
But the costs land unevenly. The employer saves recruiter hours. The candidate absorbs the anxiety, the 1am start, the unexplained rejection and the growing suspicion that nobody read anything. Nearly half of those who completed an AI interview never heard back at all.
Final thoughts
The backlash is real but I suspect it will fail, because refusing an AI interview is only an option if you can afford to refuse jobs. The 38% who walked away skew, inevitably, towards those with alternatives. Everyone else does the interview at one in the morning and hopes.
Maybe we need more Alan Sugars...?
