Ask the Employer, Because the Platform Won't Show You

Can You See the Questions on the Platform?

Not on the recorded platforms. HireVue's candidate help center says the real interview questions cannot be seen before the interview, because every candidate is given the same amount of time to review, prepare and respond to each one ¹. Radancy's help center, for the tool once called myInterview, says it in its own words: questions arrive one at a time, in order, with no option to skip one and come back ³.

The reason given is fairness rather than secrecy. HireVue says all candidates for one position get identical, employer-selected questions with the same time to prepare, and that it applies security measures to stop candidates viewing questions early ². A preview for you would be a preview only you had.

What you can see in advance is the practice round. HireVue offers three generic practice questions that can be redone without limit. It says practice recordings are never uploaded or submitted to anyone, cannot be viewed by anyone but you, and are gone once you move past the playback ¹.

Use them twice before you record anything that counts.

So decide what you want to say before the timer starts.

There is no going back, either. On Radancy's tool you can review each answer immediately after recording it, but not at the end, and once you press Next the interview moves on ³. HireVue says a candidate gets no direct access to a submitted recording through the platform. Anyone who wants to see it has to ask the hiring company, which controls the recordings ².

What Is a Public List of Reported Questions Worth?

About an hour of your time, and not much more than that. Candidates post the prompts they were asked on Glassdoor, Blind and Reddit, and for a large employer running a common platform, a search of the company name plus the tool often turns something up. What a thread gives you reliably is the shape of the round rather than the script.

The shape is what to take from a thread, and then stop reading. How many questions, roughly how long, whether the prompts want past examples or hypotheticals, whether there is a coding portion. Radancy's help center puts a typical screen on its own tool at five to ten questions with about one to three minutes an answer, and around fifteen minutes in total, as vendor averages ³.

On that tool, it is the same information without the guesswork.

What a thread cannot tell you is which of those prompts still exists. Employers are told that the public behavioural bank is the part AI coaching defeats, and that the fix is to ask about decisions the team argued over last quarter. A prompt you can find from outside is one somebody inside the company can find too.

So a leak is a smaller prize than it looks.

Assessment questions behave the same way. Whether the online assessment questions are the public ones covers what a company tag on a practice site is actually built from.

A bank taken from inside a company is a different thing from a forum post. That is internal material moved without permission, and paying for a copy puts you inside the chain rather than outside it. The risk is not only that it is stale. It is that a preparation question turns into a conduct question, which is not the ground you want to argue from.

Why Knowing the Question Does Not Win the Round

Because on a live round the follow-up decides it, and no list contains it. Live AI interviewers build the conversation as they go. Mercor says its interviewer processes a candidate's resume before the session starts, which shapes what gets asked and how deeply, and that coding problems are generated from the profile and the live conversation ⁴.

Mercor is the vendor here, describing its own system, and the market it serves is contractors for AI training work rather than employers filling their own roles ⁴. The design point survives the hedge. A question generated from your resume was on nobody's list.

Prepare the material, not the wording.

Structured interviews lean on past-behaviour questions for a reason. In one 2007 experiment, past-behaviour questions (tell me about a time when) were harder to fake than situational ones (what would you do if). The same paper reports that most candidates fake to some degree ⁵. It was an experiment with undergraduates, in person, so read it as a reason the format exists rather than proof it always holds.

The same experiment found that follow-up questioning increased faking rather than reducing it ⁵. A follow-up is not a lie detector, and nothing here says it is.

What it asks for is detail, which is the part a memorised answer does not have.

Interviewers are coached to ask which alternative you ruled out, and what would have to be true for your answer to be wrong, about your own example. Knowing the opening question does nothing for those.

Ask for Them, and Prepare the Content Instead

The message where you confirm the time is the place to ask the recruiter. Some employers send the questions to everyone, some send only the competencies they plan to ask about, and some say no. All three answers are useful, and the request itself is one line. Preparation is assumed on the other side of the table now, which is why the request reads as a scheduling question.

A barrier in the format is a separate request with its own rules. How to ask for an accommodation on an AI interview covers the wording.

Preparing hard is not cheating, and in one study the interview worked better, not worse, in the sample that had been coached. Maurer, Solamon and Lippstreu compared coached and uncoached interviewees on a situational panel interview, in a combined sample of 146 public safety job incumbents. Predictive validity and reliability were higher in the coached sample ⁶. The coaching studied was content-focused, aimed at conveying relevant material accurately rather than at moving a score by other means.

One occupation, one setting, and a comparison between samples rather than random assignment. The direction is still worth having.

So build content, not a script:

• three or four real examples you can tell in ninety seconds, each with what you did, what happened, and what you would change now
• the two or three competencies the posting keeps naming, since the questions usually come from those
• a read on the company from more than one source, which using AI to research a company before an interview covers

Then ask what the format is, while you are already asking. A one-way recording, a live AI conversation and a human panel do not reward the same preparation.

See how it works >