Your Interviewer Is Reading AI Questions. Answer the Script.

Why is a script usually good news for you?

Because the alternative is worse for you specifically, not better. In the corrected meta-analytic evidence on hiring methods, structured interviews, the same questions in the same order rated on a common scale, come out validity-ranked well above unstructured, free-flowing interviews, .42 against .19 ¹. A federal guide to running one names the reason directly: unstructured interviews are subjective, which reduces accuracy and invites legal challenge ². That subjectivity is exactly the inconsistency a candidate has no way to see or contest.

That gap exists because an unstructured conversation rewards whether you happen to click with this particular interviewer, on this particular day, more than it rewards anything about the job itself. One controlled study of unstructured interviewing, run with students predicting classmates' grades rather than managers predicting job performance, showed the mechanism at work: predictions made after an unscored, freeform chat correlated far worse with the real outcome than a single piece of prior information used alone ³.

A script is one blunt tool for removing that variable. It does not remove your own advantage or disadvantage entirely, and it is not a guarantee the team behind it is skilled. But a company running a real structured loop is choosing the format the evidence says actually distinguishes candidates on the job, not phoning it in, and that choice is worth crediting rather than reading as a lack of effort.

What tells you this is being run well, or not at all?

Watch what happens immediately after you finish an answer, not the questions themselves. A script paired with real follow-ups, a "can you say more about that" or "what would you have done differently," is a well-run structured loop: the base question is fixed for comparability, the follow-up probes are often written in advance too, and the follow-up is where the interviewer actually engages your answer rather than just checking a box that it was asked.

The questions being written or suggested by an AI tool changes almost nothing about this. What determines whether the round is diagnostic is whether a human is genuinely listening to your answer and probing it, even with a probe written in advance, and that is true whether the base question came from a rubric a person wrote five years ago or one drafted last week with an assistant.

A script with zero follow-ups, every question read and immediately followed by the next scripted line regardless of what you said, is the real yellow flag, whatever wrote the questions. It usually means nobody on that team has actually agreed what a strong answer sounds like, so nobody built the follow-up layer that finds out. That is a signal about the hiring process's maturity, not about you, and it is fair to weigh when deciding how seriously to take an eventual offer from that team.

Answer the script, then add what it didn't ask.

Answer the actual question you were asked first, directly, before adding anything else. A scripted format is comparing your answer to every other candidate's answer to the identical question, so drifting to a different topic, however impressive, costs you the comparison the format is actually built to make, even if it makes you sound broader in the moment.

Then, once you have answered it, add the one specific detail the script did not think to request: a number, a tradeoff you weighed, a mistake you caught and fixed. That detail is what a follow-up would have drawn out of you in a better-run version of the same interview, and volunteering it does the interviewer's job for them without ever sounding evasive about the question actually posed.

This works whether or not follow-ups ever arrive. If they do, you have already answered the obvious next question and the conversation moves somewhere more specific. If they never come, the interviewer's notes still carry a concrete detail instead of a generic answer indistinguishable from the next candidate's, which is the entire advantage specificity buys inside a scripted format.

What can you ask them to find out how the decision actually gets made?

Two questions surface more than a direct "is this AI-generated" ever will, and neither one puts the interviewer on the defensive. First: "What does a strong answer to that last question usually include?" A team with a real rubric can answer this specifically, often naming the exact thing they were listening for; a team without one gives you something vague, which is itself the answer, whether or not anyone intended to reveal it.

Second, near the end of the round: "What happens after this, who reviews the notes, and does someone with hiring authority read them before the next step?" A clear answer describes an actual process: a debrief, a shared scorecard, a person's name. A shrug describes the absence of any real process at all, and that absence quietly tells you roughly as much about the team as the interview itself already did.

The employer-side version of building this well names the same three requirements: identical questions, a shared rating scale written before the first candidate walks in, and fixed follow-ups, which is worth reading if you want to know exactly what a well-run version of what you just sat through is supposed to look like from the other side of the table. Neither of the two questions above requires you to name AI at all, and that is deliberate: what you are actually trying to learn is how seriously the team has thought about comparing candidates, not what software touched the question bank on the way to your screen. What you're allowed to ask about the process at all, beyond these two questions, is its own worked list.

See how it works >