What Carries Forward From an AI Round to a Human One

What Carries Forward to the Next Interviewer?

The vendors that publish their process describe three things carrying forward: the recording, a machine transcript of it, and on some platforms whatever the platform wrote about your answers. Alex's candidate FAQ says its recordings, transcripts and notes are shared with the employer's hiring team, and that humans make the hiring decisions ¹. Vervoe describes its own assessments in similar terms, telling candidates that video and audio are transcribed and provided to the employer along with the recording ².

Some platforms add a number to that package. VidCruiter's description of its pre-recorded interviews says that where AI-assisted scoring is used, reviewers receive AI-generated scores as a starting point and then apply their own judgment with built-in scorecards ³. That is a vendor stating how the feature is meant to be used. It is not evidence about how a particular reviewer weighs a score already on the screen before they press play.

Not one of those three pages is an audit.

These are companies describing their own products, which is the best public record a candidate has and still a company's account of itself. Read them for the mechanics, which a vendor has little reason to misstate, rather than for reassurance.

So the working assumption for a mixed loop is short.

• the recording of your session
• a machine transcript of what you said
• notes or a summary the platform produced from it
• on some services, a score sitting in front of the reviewer

Receiving a recording and watching one are different facts, and the second is much less certain. Whether a human ever watches an AI-scored interview is its own question, and none of these vendor pages says whether the recording is played or only read about.

One human-run technical screen lands in the same place by a different route. Karat's candidate page describes an hour run by a professional developer, and says the recruiter is sent the video recording along with a summary of performance on the questions, then decides the next step ⁴.

Why Does the Human Ask the Same Questions?

Probably because the panel wants depth, not because anyone forgot you answered. A timed round collects the version of a story you can tell inside a fixed number of minutes, and that version tends to carry the shape without the decision underneath it. The live round is where somebody can ask what you rejected, what the number was measuring, and what happened next.

A hiring loop is supposed to collect different evidence at each stage, so that no two rounds sit on the same ground. A well-built one gives every round something no other round covers, which means the recorded answers count as already collected. A repeated question then has to be doing something other than gathering them again.

Plenty of loops are not built that way.

When the same question comes back, the more useful response is the specific one. Say the same thing at more depth: the constraint you were under, the option you turned down, the thing that went wrong in week three. That is the material a summary cannot hold, and it is what a follow-up is fishing for. It is also why rehearsed STAR answers keep falling flat, since a fixed four-beat shape is the part any model can produce from a job posting in one prompt.

The recorded round supplies the other reason. A transcript is evidence a reviewer can return to, and a conversation is not. Asking again in a room is how a person checks whether the written record holds up when nobody can edit the answer.

Write Down What You Said While You Remember It

Do it the same day, because you probably cannot go back and check. HireVue's candidate help center says that after completing an interview, you have no direct access to review your own recording through the platform. A candidate who needs to see a submission has to contact the hiring company, which controls the recordings and decides whether to provide access ⁵.

Three lines per question is enough: the question as you remember it, the example you reached for, and any figure or name you said out loud. Do it before the details blur into the other applications you sent that week.

Notes made that day may be the only copy you control.

Write down what the platform told you, too. On VidCruiter, recruiters set thinking time, recording time, the number of attempts and review timers per question, and candidates can review a video before submitting it ³. Those settings explain how much room each timed answer had, and they are the mechanics that preparing for a one-way video interview tells you to check before you start.

Then read the notes next to the job description and mark the two answers you would want more room on. Those are the ones to bring detail to. Leave the rest alone, because an interviewer working from a summary picks their own places to dig, and arriving with six polished expansions is answering questions nobody asked.

Two is enough to prepare properly.

If the recording itself matters to you, ask the employer rather than the platform. HireVue's help text names the hiring company as the party that controls it and decides whether to share it, and says nothing either way about a legal right of access ⁵. A copy of whatever a vendor produced about you is a separate request, covered in getting a copy of your hiring report.

Should You Change an Answer in the Live Round?

Add to an earlier answer rather than change it, and name the change out loud if the facts move. Added detail reads as substance, because a follow-up is asking for the part a timed answer had no room for. Contradicting the recorded version reads as something else, and a machine transcript beside the interviewer makes the difference easy to see.

An honest correction is one sentence long. Something like: the number in the recorded round was a rough estimate, the real figure was closer to nine months, and here is where it came from. That names the change before anyone has to go looking for it.

Correcting yourself out loud costs a sentence.

What does not work is performing consistency. Reciting the recorded answer word for word gives a reviewer nothing the file did not already contain, and it spends the one part of the loop where you can explain yourself on a repeat. The opposite move fails the same way: treating the live round as a clean slate and telling a different story about the same project. Both answer a memory test nobody set.

What has to hold is the facts, not the wording.

On HireVue, the scorer and a person reading the file can see different things. HireVue's 2025 explainability statement says its AI relies only on what a candidate says, with no video analysis and no assessment of facial expression, body language, surroundings or tone of voice ⁶. That is the vendor's account of its own scoring model rather than an independent audit, and it says nothing about what a person notices in a recording they were sent.

Asking what the next round already has is fair, and the answer changes what you prepare. Put it to the recruiter as a scheduling question: does the panel see the recorded round, a summary of it, or neither?

See how it works >