The invitation says “AI interview.” It does not say much else.
Are you recording answers for someone to watch later? Talking to an automated interviewer? Having a transcript summarized? Those are different experiences, and the label does not tell you which one you have.
Prepare for an AI interview by confirming the format, checking the instructions, and rehearsing a clear account of work you actually did. Ask about evaluation where the employer has left it unclear. Guessing what a hidden system wants is a poor preparation plan.
Start with the invitation
Greenhouse's LinkedIn discussion of AI interviews raises a useful issue: candidates need clarity about how AI is involved. That is an employer's perspective on candidate trust, not evidence that every employer uses the same technology.
Before practicing, find out what you have been asked to do:
- Is the response recorded, typed, or conversational?
- Is there a preparation period or an answer limit?
- Are retries allowed?
- What happens if the connection fails?
- What information does the invitation give about evaluation and human review?
Do not assume you get another attempt. Do not assume there is no human reader. Work from the instructions you have, and ask about the gaps that affect your preparation.
A useful message to the recruiter
“Thanks for the invitation. Could you confirm whether this is a recorded response or a conversational interview, the preparation and answer limits, and whether retries are available? I'd also appreciate knowing how AI is used in evaluation, whether a person reviews the responses, and whom to contact for an accessibility adjustment or technical problem.”
You may not receive every detail of the assessment. You can still ask enough to understand the task and the support available.
Keep the evidence in your answer
Here is a fictional example of an answer that sounds complete while explaining very little:
“I used my leadership and analytical skills to resolve a customer issue and achieve a successful outcome.”
There is no way to tell what the issue was, what the candidate did, or what improved. Adding more job-description keywords would leave the same holes.
Now make the work visible:
“A customer could not reconcile two reports before their monthly review. I asked which totals differed, compared the date filters, and found that one export excluded the final day. I explained the mismatch and asked them to rerun the report with the same dates. They confirmed that the totals matched. I then added the date range to our export checklist so the next handoff included it.”
The example is fictional. Its value is the visible chain: problem, investigation, explanation, check. It does not depend on a claim about how an undisclosed model scores answers.
A weak answer names qualities. A credible answer shows the action. A stronger answer adds the check that supports the result and explains what the candidate learned.
If you are asked about your own use of AI, the AI mistake interview exercise helps you explain a specific check and correction rather than listing tools.
Rehearse within the actual constraint
If the invitation gives an answer limit, use that limit during practice. Read the question, pause if the format permits it, and say your answer aloud. Listen back once for missing information.
Can another person identify the problem? Can they tell which action was yours? Does the result follow from the example? Those are useful editorial checks. They are not a simulation of the employer's private scoring system.
For conversational formats, prepare for the next question too. “Why did you choose that approach?” can reveal a gap that a smooth opening hides. The follow-up question exercise gives you a way to test that depth.
Test the setup before the assessment starts
Use any employer-provided practice or equipment check. Confirm the requested browser, microphone, camera, and connection work. Find the support instructions before you need them.
Use notes only when the rules allow them. If they do, keep them to facts and brief reminders. Reading a generated paragraph does not help you explain a detail that the paragraph left out.
Keep confidential customer or employer information out of your example. You can describe a decision and its consequences without exposing names, private data, or internal documents. The confidential-work preparation guide helps you prepare a permitted version and an alternative example.
Common questions
Is a recorded video interview always scored by AI?
Recording a video does not by itself establish that AI scores it. Recording, transcription, automated questioning, and evaluation are different functions. Read the invitation and privacy information or ask the employer which functions apply.
Should I use keywords to pass an AI interview?
Use the language needed to explain your real experience accurately. Without a disclosed scoring method, claims about secret keywords are guesses. Focus on answering the question, making your contribution clear, and describing an outcome you can support.
Can interview practice predict my AI interview score?
Practice can help you notice missing evidence and improve how you explain your experience. It cannot establish your result in an employer's undisclosed assessment. A coaching score is feedback from that practice tool, not the employer's hiring decision.
How should I prepare for an AI interview?
Check whether the interview is recorded or conversational, read the employer's timing and retry instructions, test the required setup, and rehearse a real example aloud. Ask the recruiter what is evaluated and whether a person reviews the response when the invitation leaves this unclear.
Start with one real story. Use the mistake question practice page to explain what you noticed, what you corrected, and what changed. Then check your invitation again. Prepare for the interview you have actually been given.
