Interviewing
What an AI First Round Says About the Job, and What It Does Not
An AI first round is a decision about volume, made before anyone read your file, so it carries almost no information about how seriously you are being taken. It is also common now: in a survey by the hiring-software vendor Greenhouse, 63 percent of its 1,200 US-based respondents said they had been interviewed by an AI. What the format does tell you is that everyone starts from the same questions, which puts the weight on being specific.
The takeReading the format as an insult costs you more than it protects you. A first round run by software is a duller experience and a flatter field at the same time, because the question set does not change for the candidate the recruiter already likes. The signal worth reading is not the software. It is whether the employer will tell you who reviews the screen, and at what point a person enters the process.
Where Olive fits
Open a role and see what the work shows
An Olive session is not an interview, and nothing drawn from how you sound or look enters the findings: you work a real assignment with an AI assistant, think aloud or type, and a person writes or edits the findings on what happened. You are granted the same report the employer reads, free.
Assess your shortlistWhat Does an AI First Round Actually Tell You?
It tells you about the employer's applicant volume and very little about you. The choice to put software in the first round is made for a whole role, often before a single application arrives. It is a purchasing decision as much as a hiring one, argued in throughput and cost per screen.
It is also no longer unusual. Greenhouse's candidate interview report, a survey of 2,950 candidates including 1,200 US-based workers, puts the share of US job seekers who say they have been interviewed by an AI at 63 percent 3. That is a hiring-software vendor surveying its own market, with no fielding window or weighting published in the release. It counts people who have ever had the experience, not a share of interviews.
At the moment the format was chosen, it had nothing to do with you.
The employer-side article on whether to buy an AI screening tool to handle the volume prices the purchase as reading capacity: recruiter hours spent reading, days to a first human response, the share of applications a person actually opens. No single application is in that calculation.
A first round with no person in it is the ordinary shape of a large funnel now.
Does an AI Screen Make Your Odds Worse?
A randomised test found no sign that it does at the final round, and for the candidates who got that far it points the other way. Researchers assigned 37,000 applicants for a junior developer role at random 1. One track was the traditional pipeline: resume screening, then human selection. The other added a structured video interview run by AI before any human review 1.
Both tracks then faced the same final human interview, and the interviewers did not know how anyone had reached it. In that round, 54 percent of the AI-screened track passed against 34 percent from the traditional track, a difference of 20 percentage points with a standard error of 12 points 1. That error is wide enough to make the result directional rather than precise. The system also tended to advance younger applicants with less experience and fewer advanced credentials 1. That describes who this one system moved forward, and it is not an endorsement of doing it that way.
Read the result with its hedges attached. It is a preprint that has not been peer reviewed, covering one platform, one employer and one job family 1.
One study is thin ground, and it is still measured ground.
The same experiment is honest about who never got there. Of the 23,000 candidates invited to the AI interview, 24 percent completed it 2. Non-completion mixes refusal with a dead link, a lost email, a job taken elsewhere and a camera that would not work. Those who declined skewed slightly older and more experienced, and their later outcomes read from LinkedIn profiles did not differ statistically from the people who finished 2.
Inside that pipeline, though, not completing meant not advancing.
If you are weighing whether to sit one out, turning down an AI interview and still being considered is a separate question with its own answer.
Why the Format Feels Like a Slight Anyway
Because the discomfort is measured and real, even where the evidence on outcomes is not bad. A meta-analysis of technology-mediated interviews found the attitude penalty sitting in one-way and AI formats rather than in remote interviewing as such 5. Recorded one-way interviews correlated negatively with how fair candidates found the process, and with what they intended to do next 5. Interviews using AI-powered technology correlated negatively with how attractive candidates found the employer, and with those intentions 5.
Live technology-mediated interviews did not show the same penalty on fairness or intentions 5.
Those are moderator analyses resting on small numbers of studies, several with only three to five samples behind them, and the authors call the moderation partially supported 5. They are correlations with attitudes, not with outcomes. Nothing in them says a candidate is more or less likely to be hired through an AI interview. An employer running one is paying a price in how candidates see it, which is a fact about the employer's tradeoff.
A second pair of studies is worth holding next to that. One was a vignette experiment with 475 potential applicants imagining a hiring process; the other a field study of 335 people recalling a process they had been through 4. The fairness penalty for a digitalised interview stage showed up among the people imagining it, and was not significant among the people who had lived it 4.
If that pattern holds, the dread is loudest before the invitation is accepted, which is when people decide whether to bother.
That gap is not proof the worry is groundless. The vignette sample was mostly German students and young professionals, the field sample already knew its own outcome, and neither is your situation 4.
Ask Who Makes the Call After the Screen
Ask it in the reply to the invitation, because it is the one detail that changes what the format means for you. A screen that produces a transcript for a person to read is a different thing from one that closes an application on its own. It is an ordinary process question.
One live AI interviewer publishes its answer. Alex, the product formerly named Apriora, says in its candidate FAQ that it does not make final hiring decisions and does not automatically reject candidates 6. It says the recording, the transcript and the notes go to the employer's hiring team, where humans decide 6. A vendor describing its own product is not an audit, and it is not a promise about the employer either.
What a hiring team does with a transcript sits outside the vendor's control. So the question to put to the recruiter is about their process, not about the software.
Three things fit in one short message:
- who reads the recording or the transcript after the screen
- whether a person decides who moves forward
- how long the recording is kept, and who inside the company can see it
Employers get the same question from the other side. Whether an AI should conduct the first-round interview instead of a recruiter is written for the hiring team, and it tells that team to keep a named person on every advance and every rejection. Whether a human ever watches an AI-scored interview covers what that review usually looks like from where you are sitting.
The answers will not make the round warmer, and they do tell you how much of the process a person is still holding.
Common questions
Does an AI first round mean the employer is not serious about the role?
It mostly means the role drew a lot of applications. The format is chosen for a whole requisition, often before any application arrives, so it cannot be a reaction to yours. Greenhouse's candidate interview report, a survey of 2,950 candidates including 1,200 US-based workers, puts the share of US job seekers who say they have been interviewed by an AI at 63 percent, which is a vendor surveying its own market. A format that many candidates have already met is not, on its own, the mark of a careless employer. What separates them is whether they will tell you who reviews the screen.
Will the AI decide on its own whether I move forward?
Two answers are needed here, one from the product and one from the employer. One live AI interviewer, Alex, says in its candidate FAQ that it does not make final hiring decisions and does not automatically reject candidates, and that the recording, the transcript and the notes go to the employer's hiring team, where humans decide. The company wrote that about its own product, which is not an audit, and it says nothing about what another employer does with another tool. Ask the recruiter who reads the screen and who decides after it.
Does going through an AI screen hurt my chances compared with a human first round?
A randomised test bearing on that question found no harm at the final round, imprecisely. Among 37,000 applicants for a junior developer role, 54 percent of the track that went through an AI structured video interview passed the final human interview, against 34 percent of the track screened the traditional way. It is a preprint that has not been peer reviewed, from one platform and one job family, and the standard error of 12 points against a 20 point effect is wide. Directionally, candidates who reached the final round through the AI stage did better there, not worse, though in that pipeline those who did not complete the AI interview did not advance.
Is it reasonable to ask for a person instead?
Yes, and asking who reviews the screen is more useful than asking for the screen to be replaced. A live conversation at some stage is a reasonable thing to ask for, and it is a different request from declining the round. In the one large experiment on this question, candidates who did not complete the AI interview did not advance in that pipeline, so treat skipping one as a decision rather than a default.
What should I do differently in an AI first round?
Be specific, because the format takes away the things that usually carry a first conversation. Everyone starts from the same questions, so an answer has to stand on its own without a person reading the room. Name the situation, what you did, what happened, and how you know it worked. Vague answers sink applications in any format, and a structured screen makes them easier to see.
References
- 1. Better Together: Quantifying the Benefits of AI-Assisted Recruitment arxiv.org The randomised field experiment behind the pass-rate comparison in this article: 37,000 applicants, 54 percent of the AI-assisted track passing the final human interview against 34 percent, with the preprint status and the wide standard error stated alongside it.
- 2. Better Together: Quantifying the Benefits of AI-Assisted Recruitment arxiv.org The completion figure this article uses for who never reached the AI interview, and the finding that people who declined were slightly older and more experienced with no statistical difference in their later outcomes.
- 3. 63% of Job Seekers Have Faced an AI Interview. Most Haven't Had a Good One Yet greenhouse.com Supports the claim that an AI interview is now the common case for a US job seeker, with the survey's sample sizes and its status as a vendor's own survey stated in the sentence that uses it.
- 4. Applicant Reactions to Digital Selection Methods: A Signaling Perspective on Innovativeness and Procedural Justice link.springer.com Supports the finding that the fairness penalty at a digitalised interview stage appeared among people imagining a process and was not significant among people recalling a real one, with the sample descriptions this article names.
- 5. Technology-mediated interviews: meta-analyses on interview ratings and job applicants' responses link.springer.com Supports the claim that the attitude penalty sits with one-way and AI formats rather than with live technology-mediated interviews, and the caution that these are correlations with attitudes drawn from small numbers of samples.
- 6. Candidate FAQs alex.com The vendor's own statement that it does not make final hiring decisions or automatically reject candidates, and that the recording, transcript and notes go to the employer's hiring team.
6 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.