Job seekers report being interviewed by AI instead of humans

[NightOwl]

British Job Hunters Turning Up To Find Their Interviewer Is An Ai Robot 13531971news.sky.com

More candidates are encountering AI interviewers in hiring processes, often without clear disclosure beforehand. Companies are using AI to screen applicants at scale, but this raises concerns around fairness, transparency, and candidate experience. It also signals how deeply AI is embedding itself into everyday workflows beyond just backend automation

[RayOfLight]

Would be funny if it werent true!

[Matt_81]

This feels like a pretty big shift in how hiring works
Could introduce bias in ways that are harder to detect

[Lucy05]

Also removes a lot of human judgement from early stages

[One-One-Five]

Some candidates will adapt quickly, others will be filtered out unfairly
Transparency about AI use should probably be mandatory

[AJStyles92]

Would be funny if it werent true!

Cheers for that. Every time without fail.

Good stuff.

The useful stuff is harder to spot because there is so much noise around it

[Chris27]

One concern that does not get enough attention is accessibility. A human interviewer can adapt when someone communicates differently, whereas an automated system may assume there is one correct way to sound confident, engaged or professional.

Consider someone who gives excellent written answers but has anxiety during video calls. Or someone with a hearing difficulty who needs captions. If the system quietly treats hesitation, eye movement or speech patterns as indicators of suitability, candidates may be penalised for characteristics that have little to do with their ability to perform the job.

That does not mean AI interviews are automatically discriminatory, but it does mean companies need to validate these systems properly rather than assuming an algorithm is neutral because it is an algorithm. The burden should be on the employer to demonstrate that the process is fair.

And yes, telling people that an AI is evaluating them after they have already completed the interview is a particularly poor way to build trust.

[Inland Renegade]

The funny thing is that companies are trying to make interviews more efficient while potentially making candidates spend even more time preparing for them. If people start learning exactly which phrases, facial expressions and answer structures supposedly impress the model, the process becomes another optimisation game.

That could create a strange feedback loop. Candidates learn how the system scores them, coaching companies sell courses on beating the system, and employers then change the model to detect the coaching. Before long the job interview has turned into competitive exam preparation with a webcam attached.

The better use of AI would be helping the human recruiter rather than replacing the human interaction completely. It could summarise answers, flag relevant experience and suggest follow-up questions, while the recruiter decides whether the person is actually a good fit.

Technology should reduce pointless recruitment admin, not make applicants feel like they are trying to pass a secret test designed by a spreadsheet.

[FrostDrifter]

There is a practical argument for AI interviews, especially when a company receives thousands of applications. Asking every candidate the same initial questions can save recruiters a huge amount of time, and candidates might get a quicker response instead of waiting three weeks for an email that says nothing.

The problem is when the tool becomes a black box. If someone gets rejected after an automated interview, they should at least have some idea what was assessed. Was it relevant experience, the substance of the answers, communication skills, or some mysterious model-generated score? Without that information, there is no meaningful way to challenge a bad result.

A sensible compromise would be AI for the administrative first pass, followed by a human review for serious candidates. Let the machine handle the repetitive questions, but do not let it become judge, jury and recruitment department all at once. Otherwise we will eventually need an AI to interview the AI interviewer.

[Diana63]

The disclosure part is what bothers me most. Using AI to handle the first stage of an interview is one thing, but candidates should know before they start talking to a machine. Imagine spending an hour preparing thoughtful answers only to discover the other side is an automated assessment that may be scoring tone, wording and pauses as well as the actual content.

There is also a fairness question here. A good human interviewer can notice that someone is nervous, has a speech impediment, is communicating in a second language or simply needs a moment to think. An automated system may interpret the same behaviour as a negative signal. That is a pretty big consequence for something as subjective as an interview.

AI can certainly remove some repetitive work from recruitment, but there should be a human involved where the decision actually matters. Otherwise companies risk optimising the hiring process for efficiency rather than finding the best people.

[Serpent23]

There is a difference between automation and impersonation that I think gets lost in this debate. An automated questionnaire can be perfectly reasonable if it is clearly presented as one. Pretending the candidate is having a normal conversation when every response is being processed by a model feels much more questionable.

Recruitment already has enough uncertainty without adding another invisible layer. Candidates are often wondering whether a human actually read their CV, whether the salary range is real and whether the advertised position is still open. Finding out that the interview itself was automated can make the whole process feel even more detached.

Personally, I would be much happier with a message saying, "This first-stage assessment is automated and takes around 15 minutes. Your answers will be reviewed alongside your application." At least then everyone knows the rules before agreeing to play.

That small amount of transparency would probably prevent a lot of the backlash. People are generally more accepting of automation when they know exactly where it is being used.

[Static Estuary]

There is another side to this that is worth considering. Human interviews are hardly a perfect gold standard. People have unconscious biases, favourite personalities and completely different ideas about what confidence looks like. One interviewer might love a candidate while another dislikes them within thirty seconds for reasons that have nothing to do with the job.

A carefully designed automated process could potentially make certain parts of hiring more consistent. If every candidate gets the same questions and the same basic criteria, that might remove some arbitrary variation. The catch is that the criteria themselves can still be biased, and a model trained on historical hiring decisions can reproduce the preferences of the humans who made those decisions.

So replacing humans with AI does not automatically remove bias. It can simply move the bias into a less visible place. That is arguably more dangerous because candidates cannot see the assumptions being made about them.

Used as a tool with auditing and human oversight, AI could improve recruitment. Used as an unquestionable gatekeeper, it is just bureaucracy with a GPU.

[DigitalNomad54]

The most sensible question is probably not whether AI interviews are good or bad, but what part of the process they are actually being used for. Asking basic availability and experience questions is quite different from deciding whether someone's personality is suitable for a team based on a video recording.

For example, a system could collect structured answers and give the recruiter a summary of relevant experience. That seems fairly harmless and potentially useful. Having another system infer that someone is "not leadership material" because they paused for two seconds before answering is a completely different proposition.

There should also be a straightforward way for candidates to request human review, particularly when an automated assessment produces a rejection. If a company is confident that its system works, allowing a reasonable appeal should not be a problem.

Otherwise recruitment starts to resemble customer service menus where you spend ten minutes saying "representative" to a robot. The only difference is that this time the robot is deciding whether you get paid next month.