Prompt AI to Turn a Job Posting Into Interview Questions
A prompt that turns a job posting into role-specific interview questions, plus how to avoid the generic list every other candidate has already seen.
Prompt AI to Turn a Job Posting Into Interview Questions
Paste the full job posting and ask: “Generate 10 interview questions specific to this role, based on the actual responsibilities and requirements listed, not generic interview questions. For each, note what a strong answer would demonstrate. Include at least 2 questions that would expose a candidate who is overstating their experience with something specific in this posting.” The third sentence is what separates this from a list any candidate has already rehearsed against ChatGPT themselves.
Why generic question banks stopped working
“Tell me about a time you showed leadership” has been asked so often that a prepared answer exists for it before the interview starts. Candidates now routinely rehearse against AI-generated question banks, which makes the generic list actively worse than useless: it screens for preparation, not fit. A prompt anchored to the specific responsibilities in your actual posting produces questions a candidate cannot have pre-written a polished answer for, because the questions depend on details only your posting contains.
The full prompt, step by step
Paste the complete job posting, including the parts you might consider boilerplate, since specific tools, metrics, or scope details often live there.
Ask for questions tied to specific lines in the posting, not the role title in general.
Request a “what a strong answer demonstrates” note for each question, so you know what you are actually listening for before the interview, not during it.
Explicitly ask for at least two questions designed to expose overstated experience on something the posting calls for, such as a specific tool or years of experience with a technique.
If the role is technical, ask for one question that requires walking through reasoning out loud rather than stating a conclusion, which is much harder to fake.
An example of the difference this makes
A generic question for a “senior backend engineer” posting: “Describe your experience with databases.” A posting-specific question generated from a listing that mentions “migrated a legacy system from MySQL to Postgres under load”: “Walk me through how you would plan a live migration from MySQL to Postgres for a system that cannot have downtime. What would you check first, and what would make you postpone the migration.” The second question cannot be answered from a memorized template, because it demands specific reasoning the posting itself set up.
The bias caution this comes with
AI-generated interview questions can quietly encode bias if you are not watching for it, especially questions that reward a specific communication style or cultural reference over the actual skill being assessed. Review every generated question for whether it is testing the job requirement or testing familiarity with something unrelated to it. Our piece on whether your hiring process needs a bias audit covers this in more depth if AI is touching more of your hiring process than just question generation.
Pairing this with the rest of your hiring process
Interview questions are one piece of a hiring process that AI can help with more broadly. If a posting is bringing in a flood of applicants at once, using AI to handle a seasonal hiring spike covers the volume side. Once someone is hired, the same posting-specific, evidence-anchored approach applies to prompting AI to write a performance review, which runs into the same generic-versus-specific problem from the other direction.
FAQ
How many questions should I actually ask in one interview?
Generating 10 gives you room to pick the 4 or 5 that fit your actual interview length and drop any that turned out weaker than expected, not a target to ask all of them.
Can I use this for a role I am writing the posting for myself?
Yes, and it doubles as a check on the posting itself: if the model cannot generate a specific question from a line in your draft, that line is probably too vague to attract or screen candidates well either.
Should candidates know the questions came from AI?
There is no obligation to disclose your prep process, but the questions should be ones you understand well enough to evaluate the answer to, regardless of where they came from.
How did this land?
About the author
Developer Advocate
Steve builds something with Swarmz every week and writes up what worked, what broke, and what he'd do differently. Tutorials and hands-on guides are his lane.
